Interdependence among productive activities: Implications for exploration and exploitation

JBR-08136; No of Pages 12 Journal of Business Research xxx (2014) xxx–xxx

Contents lists available at ScienceDirect

Journal of Business Research

Interdependence among productive activities: Implications for exploration and exploitation Corrado Gatti ⁎, Loredana Volpe, Gianluca Vagnani Sapienza, University of Rome, Faculty of Economics, Department of Management, 9 Castro Laurenziano St., 00161 Roma, Italy

a r t i c l e

i n f o

Article history: Received 22 April 2013 Received in revised form 14 January 2014 Accepted 30 July 2014 Available online xxxx Keywords: Exploration Exploitation Interdependence Decomposability Firm's long-run financial performance

a b s t r a c t The objective of this study is to explore how the level of interdependence and that of decomposability at the industry level moderate the contribution of exploration and exploitation to firms' long-run financial performance. We employ patent data to measure interdependence and decomposability and computer-assisted content analysis to derive firms' orientations toward exploration and exploitation. We also introduce statistical techniques to control for biases in estimates induced by potential sources of endogeneity. Based on our analysis, in industries that exhibit high levels of interdependence and low levels of decomposability, exploration becomes more necessary to improve firms' long-run financial performance. On the other hand, in industries that exhibit more limited levels of interdependence and high levels of decomposability, exploitation becomes more beneficial to firms' long-run financial performance. We hope our findings will stimulate future research on a number of distinct but related issues, including exploration, exploitation, interdependence, and decomposability, and thus contribute to improve our understanding of organizational success. © 2014 Elsevier Inc. All rights reserved.

1. Introduction When is it beneficial for firms to invest either in the exploration of new combinations of productive activities or in the exploitation of existing ones in order to achieve greater long-run performance? This question has inspired a wide range of studies in different domains, including organizational learning and strategy (e.g., Levinthal & March, 1993; March, 1991), innovation (e.g., Danneels, 2002), the search for new technologies (e.g., Fleming, 2001), organizational design (e.g., Tushman & O'Reilly, 1996) and entrepreneurship (e.g., Shane & Venkataraman, 2000). Despite such extensive literature, the posed question still raises a dilemma. On the one hand, organizations must invest in exploratory search in order to innovate, while also considering that exploration exposes the firm to greater risks of failure and increases the costs of integrating new processes and products with wellestablished ones. On the other hand, organizations must invest in exploitation in order to limit search costs, and more easily reap the benefits of already deployed combinations. Yet, an organization that focuses more on exploitation may suffer an expanded risk of obsolescence and declining benefits in the long-run (Levinthal & March, 1993; Levitt & March, 1988). Therefore, how can managers know whether paying more attention toward exploration or exploitation is conducive to higher firms' long-run performance?

⁎ Corresponding author. E-mail addresses: [email protected] (C. Gatti), [email protected] (L. Volpe), [email protected] (G. Vagnani).

An important stream of the extant literature on complex adaptive systems begins to unfold this issue by illustrating how the long-run performance effects of exploration and exploitation are influenced by the interdependencies that characterize the environment in which organizations operate (Lenox, Rockart, & Lewin, 2006, 2007; Levinthal, 1997; Rivkin & Siggelkow, 2007). Interdependencies exist whenever the value of conducting a given productive activity or set of productive activities depends on how an organization conducts other activities (Lenox, Rockart, & Lewin, 2010). Activities that are subject to interdependencies typically include those related to the introduction of new aspects of organizational forms, new features in the manufacturing processes and/or new product characteristics (Lenox et al., 2010: 121).1 Using simulation methods, prior theoretical studies have captured interdependencies and show that the higher the level of interdependence, the greater the number of sub-optimal combinations of productive activities that firms may encounter in their search activities, and the greater the risk for firms of being trapped into inferior combinations of productive activities (Levinthal, 1997). If organizations wish to cope effectively with interdependencies, they must broaden their search and focus more on exploration and less on exploitation. In particular, a

1 The presence of interdependencies among productive activities implies, for example, that, in semiconductor industry, changing the technology for a mask will either improve or worsen the performance of technologies used by the firm to align the mask with the semiconductor. Another example is the introduction of a computer-aided engineering system that may require changes in other activities for the production process to work properly (e.g., the introduction of computer-aided manufacturing workstations and/or flexible manufacturing systems).

http://dx.doi.org/10.1016/j.jbusres.2014.07.011 0148-2963/© 2014 Elsevier Inc. All rights reserved.

Please cite this article as: Gatti, C., et al., Interdependence among productive activities: Implications for exploration and exploitation, Journal of Business Research (2014), http://dx.doi.org/10.1016/j.jbusres.2014.07.011

2

C. Gatti et al. / Journal of Business Research xxx (2014) xxx–xxx

greater focus on exploration allows organizations to extend the number of alternatives and trajectories within their reach, limits the risk of becoming prematurely locked into inferior alternatives, and mitigates the risk that changes in interdependencies will rapidly render currently deployed combinations obsolete (Levinthal, 1997; Rivkin & Siggelkow, 2006). Although both exploration and exploitation are needed to improve long-run performance, high levels of interdependence may thus render exploration more beneficial than exploitation to improve long-run financial performance. Extant theoretical studies have then illustrated how not only the level of interdependence but also the structure of such interdependencies (i.e. the underlying pattern of interactions among productive activities at the industry level) may influence the long-term benefits of an organization's exploratory and exploitative efforts. In particular, we refer to decomposability as the distribution pattern of interdependencies (Ethiraj, Levinthal, & Roy, 2008; Rivkin & Siggelkow, 2007; Yayavaram & Ahuja, 2008; Zhou, 2013). A set of productive activities is highly decomposable when interdependencies among individual groups of activities are limited in number and weak in intensity as compared to those occurring within each group, which are many in number and strong in intensity (Simon, 1962, 2002). Scholars have emphasized the ambivalent implications of decomposability on the benefits of exploration and exploitation to long-run performance. It has been argued that the tendency of productive activities to be decomposable, on the one hand, renders exploration essential to improve long-run performance because it expands the number of sub-optimal alternatives and therefore the complexity faced by organizations in their adaptive processes (Rivkin & Siggelkow, 2007). On the other hand, the higher the level of decomposability, the easier it is for organizations to observe, understand and even cope with interdependencies among productive activities, thanks to reduced design and cognitive complexity, which makes exploration both less effective than exploitation for firms' long-run performance and more vulnerable to imitation (Ethiraj & Levinthal, 2004; Frenken, Marengo, & Valente, 1999). Notwithstanding such contributions, extant literature on interdependencies among productive activities remains largely theoretical or still relies on computer simulation. More empirical evidence is therefore required on the moderating effect of interdependencies, and on how it conditions a firm's long-term financial performance. Note that testing the propositions developed within extant theoretical studies represents an important task for empirical research (Davis, Eisenhardt, & Bingham, 2007) and, as observed by Lenox et al. (2007), developing good ways to measure interdependencies is one of its biggest challenges. In this study, we contribute to the literature by providing a largescale empirical test of the basic theoretical relationships between exploration, exploitation and long-run performance, controlling for the moderating effect of interdependence. These relationships are analyzed in a longitudinal panel research design that covers the years 1989–2008 for 460 firms included in the 1989 Standard & Poor's 500 (S&P 500) and 400 (S&P 400) indexes. We clarify the moderating role of interdependencies at the industry level, showing whether such interdependence can magnify, attenuate, or even reverse the effects of exploration and exploitation on firms' long-term financial performance. In that, we provide empirical ground to extant theoretical studies that relate interdependencies, exploration, exploitation and firms' long-run performance. In addition, we develop reliable measures of interdependence based on public available data. Furthermore, the proposed measures vary not only across industries but also over time. The availability of time-varying measures of interdependence and decomposability makes it possible to control for the causality relationships among our main independent variables and allows us to use econometrical methods able to control for potential distortions of estimates induced by endogeneity and unobserved heterogeneity (Hamilton & Nickerson, 2003). Overall, this study provides a fuller understanding of how organizations may successfully respond to multiple environmental conditions

through pursuing exploratory and exploitative search strategies. Interdependence matters, and may lead to diverse performance outcomes for exploration and exploitation activities. Interdependence should thus be placed among the most central concepts in the business field. 2. Literature review and hypothesis As observed by March (1991), the relationship between the exploration of new possibilities and the exploitation of old certainties is crucial to studying firm performance over the long-run. Exploration includes things captured by terms such as “search, variation, risk taking, experimentation, play, flexibility, discovery, innovation.” Exploitation includes such things as “refinement, choice, production, efficiency, selection, implementation, execution” (p. 71). Interpreted in March's spirit, exploration refers to distant, systemwide search. In such broad terms, exploration is associated with path breaking, improvisation, autonomy and chaos, and emerging markets and technologies (March, 2006). In essence, an organization enacts exploration by broadly spanning numerous and unprecedented combinations of individual activities. Involving a system-wide perspective, a wider time commitment and a broader space horizon (March, 2008), exploration helps fight organizational myopia and competency traps (Levitt & March, 1988) and extends a firm's search beyond the neighborhood of currently known alternatives (Abernathy & Clark, 1985; Fleming, 2001; Rosenkopf & Nerkar, 2001). In addition, exploration stimulates the development of new skills and capabilities, which reduces the risk of becoming obsolete (Leonard-Barton, 1992). It also favors experimentation with expanded sets of opportunities lying beyond local alternatives, which leads to the introduction of new products and production processes, the creation or access to new markets, and the regeneration of consumer value (Fleming & Sorenson, 2001). These outcomes are expected to contribute positively to the organization's long-run financial performance (Lewin, Long, & Caroll, 1999). Exploitation involves the introduction of new combinations that grow out of the old by continuous adjustments, in small steps, and implies a relatively restricted search for alternatives that complement an existing technology. Exploitation thus allows an organization to reduce the likelihood of errors and false starts, and facilitates the development of routines, making search more reliable (Levinthal & March, 1981). Moreover, it favors the use of accumulated knowledge, which boosts the firm's ability to introduce new products or new productive processes that in many important ways may be not apparent to less experienced organizations (Katila & Ahuja, 2002). By reducing variety, increasing efficiency in current operations, and enhancing adaptation within current markets and with existing customers, exploitation matters for firms' long-run financial performance (Lavie, Stettner, & Tushman, 2010). 2.1. The moderating role of interdependence on the long-term financial performance effects of exploration and exploitation Both exploration and exploitation are necessary for (March, 1991) and influence firms' long-run financial performance (Jansen, Van den Bosch, & Volberda, 2006). However, the net benefits of these search activities depend at least on two different components: search costs and payoffs stemming from such activities (Levinthal & March, 1993; March, 1991). Extant complexity literature has also examined how the presence of interdependence that characterize the environment in which organizations operate influences the net benefits of exploration and exploitation (Ethiraj & Levinthal, 2004; Lenox et al., 2006, 2007; Levinthal, 1997; Rivkin & Siggelkow, 2007). Interdependence manifests itself when the value of a given activity is dependent upon the characteristics of the activity itself as well as upon the characteristics of other productive activities. Interdependence is here conceived to be an industry-level attribute that is faced by all the firms belonging to a given industry rather than chosen by the firm itself. We instead treat

Please cite this article as: Gatti, C., et al., Interdependence among productive activities: Implications for exploration and exploitation, Journal of Business Research (2014), http://dx.doi.org/10.1016/j.jbusres.2014.07.011

C. Gatti et al. / Journal of Business Research xxx (2014) xxx–xxx

exploration and exploitation activities as firm-level attributes that derive from a firm's choices. 2.1.1. The level of interdependence In the presence of extended levels of interdependence at the industry level, firms face barriers to search and often struggle to discover valuable configurations of productive activities, even via imitation processes (Lenox et al., 2010). In particular, greater levels of interdependence at the industry level induce a higher potential for interdependence among productive activities at the firm level (Lenox et al., 2006, 2007). That is, in searching for new combinations of productive activities the effectiveness of each decision by firms to change one activity depends on how other activities are conducted (Lenox et al., 2007). Thus, high levels of interdependence at the industry level increase trade-offs between firms' productive activities, enhance any constraints imposed by the initially developed combination on subsequent improvements (Lenox et al., 2007), and also increase the dependence of current and future organizational efforts on past choices (see also Levinthal, 1997). In addition, extensive levels of interdependence boost the number of local inferior optima and potentially unprofitable trajectories (Lenox et al., 2007). Therefore, the higher the level of interdependence, the greater the risk for an organization of becoming prematurely locked into a neighboring combination of productive activities that has inferior long-run potential. Scholars have acknowledged that extensive levels of interdependence are likely to expand the net benefits of pursuing more exploration and its contribution to firms' long-run performance (e.g., see Levinthal, 1997; Lenox et al., 2007). The exploration of new possibilities allows organizations to attenuate their dependence on their currently deployed combinations of productive activities and related path of improvements. It mitigates the firm's risk of becoming prematurely locked into inferior alternatives, extends the number of combinations within the firm's reach, and helps drive the organization into the basins of attractions of superior combinations of productive activities (Levinthal, 1997). Hence, although exploration is a high-cost search, it permits the organization to overcome and profit from the challenges that greater levels of interdependence impose on organizational adaptation. In addition, extensive interdependencies hamper imitation (Lenox et al., 2007; Rivkin, 2000), and increase the negative effect of errors in the reproduction of one or more elementary components by competitors (Rivkin, 2000). In that, interdependencies allow a firm that has discovered and selected a new combination of productive activities to more easily retain its value, which further contributes to enlarging the net benefits of exploration, and strengthens the contribution of such an activity to firms' long-run financial performance. Conversely, more extended levels of interdependence at the industry level are likely to make greater levels of exploitation less beneficial to firms' long-run financial performance. Exploitation sustains the refinement and improvement of the already deployed technology. As such, costs associated to exploitative activities are comparatively lower than those required by exploration (Jansen et al., 2006). Yet, decisions that focus on exploitation run out of improvement possibilities quickly when productive activities are highly interdependent (Rivkin & Siggelkow, 2007). Moreover, the proliferation of multiple local inferior optima and the emergence of path dependent patterns of adaptation increase the risk for organizations of being prematurely locked into a neighboring technological solution currently deployed or settling into a trajectory that has inferior long-run potential (Levinthal, 1997). As noted by Kauffman (1993: 53), “inexorably in landscapes with extensive levels of interdependency, adaptive walks terminate on poorer solutions”. Such a premature lock-in can stop the search for good alternatives at early suboptimal solutions, and expose organizations to the threat of higher-potential laggards and new entrants. These potential competitors may take over the scene, eventually introducing new innovative combinations via exploratory search, and thus overtaking those organizations that are focusing on the refinement of

3

already deployed combinations (Lenox et al., 2007; Rivkin, 2000). Both exploration and exploitation are necessary for firm performance, but a greater potential for interdependence among firms' activities makes exploration even more essential. H1. The level of interdependence (a) positively moderates the relationship between exploration and firms' long-run financial performance, and (b) negatively moderates the relationship between exploitation and firms' long-run financial performance.

2.1.2. The level of decomposability Given their number, interdependencies can have different distribution patterns (or structures) (Ghemawat & Levinthal, 2008; Rivkin & Siggelkow, 2007). The structure of interdependencies is gradual in nature, with most actual interdependencies lying along an ideal continuum between the extremes of perfect decomposability and total non-decomposability (Simon, 1962, 2002). In a decomposable structure, interdependencies among individual clusters of activities are limited in number and weak in intensity as compared to those occurring within each cluster, which are many in number and strong in intensity. Conversely, in a non-decomposable structure, interdependencies are such that each component is linked with almost equal strength to all other components (Buenstorf, 2005; Frenken et al., 1999; Langlois, 2002). When the level of decomposability at the industry level is high, most interdependencies occur within specific groups of productive activities, while a few occur across groups, and the latter are also weak in intensity. As a consequence, organizations potentially face lower design complexity in their search activities (Ethiraj & Levinthal, 2004). Consider the example of building a racing yacht by the New Zealand team (Iansiti & MacCormack, 1996). Despite the extended number of interdependencies, a racing yacht can be perceived as being composed of four main blocks: the hull, the keel, the mast, and the sails. The team focused on the design of the hull, applying only very simple keel variations and ignoring the mast and the sails. Only when a robust design for the hull and the (simple) keel evolved, did the team move from designing prototypes (that were limited to some aspects of the overall system) to the design of a real yacht (see also Baumann & Siggelkow, 2012). As the example suggests, the team considered many of the interdependencies as uninfluenced, which greatly facilitated the search for a new racing yacht. Extensive levels of decomposability attenuate trade-offs insisting among firms' productive activities, and reduce the risk for an organization of being prematurely trapped into inferior combinations of productive activities (Lenox et al., 2007). The greater the level of decomposability, the less is exploration necessary for firms' long-run financial performance. Although firms with greater orientation toward exploration activities are those that are more likely to find a global optimum, such organizations face a higher risk to be displaced by competitors that rely on simpler and more local strategies. Indeed, benefitting from a reduced design complexity, competitors that conversely focus more on exploitation may be locked into local optima, but reach them relatively quickly and with lower expenses (Frenken et al., 1999). In addition, thanks to a reduced barrier to search, competitors may have wider opportunities and more time to rediscover a firm's configuration of productive activities via local search, and to copy its successes (Rivkin & Siggelkow, 2007). Thus with extensive decomposability and over the long-run, a firm investing more resources in high-cost exploratory search strategies may experience a major risk of being overcome by the competitive forces (Zahra, 1996). On the flip side, high levels of decomposability at the industry level expand the net benefits of pursuing more exploitation at the firm level. Firms can indeed decompose the search space into smaller subspaces, search locally in these subspaces for improved combinations of productive activities independently of each other, and determine a superior configuration of activities by combining improved solutions

Please cite this article as: Gatti, C., et al., Interdependence among productive activities: Implications for exploration and exploitation, Journal of Business Research (2014), http://dx.doi.org/10.1016/j.jbusres.2014.07.011

4

C. Gatti et al. / Journal of Business Research xxx (2014) xxx–xxx

found in each subspace. Greater levels of exploitation allow then a firm to continuously refine already deployed combinations of productive activities, and introduce improved new products and/or new manufacturing processes without increasing the risk of premature lock-in (Brusoni, Marengo, Prencipe, & Valente, 2007; Frenken, 2006). At the same time, firms continue to benefit from knowledge accumulation (Stuart & Podolny, 1996), limit the cost of combining diverse sources of knowledge (Lee & Allen, 1982), and minimize the risk of failure (Levinthal & March, 1981). The above discussion suggests the following hypothesis. H2. The level of decomposability (a) negatively moderates the relationship between exploration and firms' long-run financial performance, and (b) positively moderates the relationship between exploitation and firms' long-run financial performance.

2.2. Previous empirical studies In evaluating the differential performance implications of exploration and exploitation activities, previous studies have concentrated on environmental aspects such as dynamism, munificence, competitiveness, and appropriability of technological innovations (for a review see Lavie et al., 2010). Yet, no study has directly explored both how the presence of interdependence among firms' activities and the structure of such interdependencies moderate the relationship between exploration, exploitation and firms' long-run financial performance. Some related empirical studies have only considered how interdependencies may affect organizational adaptation. More precisely, researchers have focused on interdependencies as an industry-level attribute. In this vein, Lenox et al. (2010) provide a cross-sectional study showing the implications of different levels of interdependencies on the distribution of profits across firms. A number of studies also shows that firms developing complex technologies face higher risks of failure (Singh, 1997) or that interdependence among the components of a complex product shapes a firm's allocation of inventive efforts (Ethiraj, 2007). Yet, other contributions explain that interdependence across international operations affects subsidiary performance in multinational firms (Subramaniam & Watson, 2006), and coupling between technological components affects the value of scientific knowledge in inventive activities (Fleming & Sorenson, 2004). Empirical studies on the structure of interdependencies are rare, and concentrate either on the implications of patterns on the degree of organizational divisionalization and hierarchy within firms (Zhou, 2013), or they consider the structure of interdependencies as a result of a specific set of a firm's choices. Particularly, Yayavaram and Ahuja (2008) argue that firms' choices of combining and considering disparate components as interdependent on each other limit exploratory activities, and make effective recombination of any newly identified elements into successful inventions more complex and difficult. Conversely, consideration of many activities as extremely decomposable emphasizes the benefits of exploration, but may provide no integration mechanisms to link the results of this exploration across clusters, thus again limiting the likelihood of successful invention (Yayavaram & Ahuja, 2008). These studies provide guidance for measuring interdependencies and their structure. 3. Method Four main data sources are enlisted in this study: Compustat, the U.S. Patent and Trademark Office, LexisNexis, and Thompson Financial. Firstly, the study refers to Compustat in acquiring financial and accounting information. The original sample is composed of 460 companies listed in the 1989 S&P 500 large-cap and S&P 400 mid-cap indexes whose primary Standard Industrial Classification (SIC) codes range from 2000 to 3999 and from 7370 to 7374. These SIC codes include both traditional manufacturing industries and information technology industries for

which our measures of exploration, exploitation, and interdependence are particularly significant (Chen, 2008; Lenox et al., 2010; Uotila, Maula, Keil, & Zahra, 2009). In must be further noted that combining large and non-large organizations in our sample allows our estimates to consider organizations whose behaviors and performance have been deemed to be less (large)/more (non-large) affected by environmental conditions (Miller, 1987). As a robustness check, we test our results for each group of firms and observe consistent estimates. We use the 20 years closest to 1989 (1989–2008) as the time period for this study in order to capture sufficient time variations in the variables of interest (Arellano, 2003). Such time frame also permits comparability with other studies addressing the effectiveness of exploration and exploitation activities in a longitudinal research design (e.g., Benner & Tushman, 2002; Rosenkopf & Nerkar, 2001). Limitation of the records to only those allowing calculation of our dependent and independent variables leads to an initial dataset of 7026 observations. Secondly, from the U.S. Patent and Trademark Office, we collect all available information attached to utility patents whose application date ranges from 1979 to 2008. Despite limitations, patent data have many important features (Griliches, 1990; Patel & Pavitt, 1997). Particularly, they allow us to directly capture interdependencies related to the introduction of new features in the manufacturing processes and/or new product characteristics (Fleming & Sorenson, 2004; Sorenson, Rivkin, & Fleming, 2006). In total, we analyze approximately 4 million patents and almost 20 million co-occurrences of subclasses. Lastly, we refer to LexisNexis and Thompson Financial databases to measure exploration and exploitation orientations. For each organization included in the sample, we collect archival data available in annual reports, and specifically within the most widely read sections of the latter (Courtis, 1982; Previts, Bricker, Robinson, & Young, 1994), namely president's letters to stockholders and management's discussions and analyses of financial results. Such documents have been shown to constitute a forum for managers to discuss themes that are important to the firm (Osborne, Stubbart, & Ramaprasad, 2001) and its performance (Abrahamson & Amir, 1996; Bowman, 1984; Michalisin, 2001). Annual reports that were not available in text format were transformed into electronic files. 3.1. Independent variables 3.1.1. Interdependence measures Several techniques have already been used in the literature to measure both the level of interdependencies among productive activities and their distribution pattern. Using survey data, most studies have measured interdependencies at the firm level (e.g., Cassiman & Veugelers, 2006; Ichniowski & Shaw, 1999) or that generally influence performance for all firms within an industry or across industries (Lenox et al., 2006, 2010). To measure decomposability, some recent contributions have instead resorted to input–output (IO) tables (Zhou, 2013). Notwithstanding the formal correctness of such approaches, a critical issue in both using cross-sectional and IO research design is providing good confidence for causal inference. Indeed, studies under these approaches are conducted at one point in time, and therefore do not account for the possibility that the performance implications of interdependence observed in one period may differ in the long-run. This may happen especially when changes in the dependent variable are expected to occur at different pace after the intervention of the independent variable (as it is for the impact of search strategies on firms' long-run financial performance). Moreover, in a cross sectional design, controlling for endogeneity may not be possible. In this study, we construct measures of interdependence (and decomposability) that vary not only across industries but also over time. In measuring the level of interdependence at the industry level, we approximate the set of components (e.g. devices, tools, and knowledge) used by organizations in activities related to the introduction of new products and/or new productive processes through subclasses

Please cite this article as: Gatti, C., et al., Interdependence among productive activities: Implications for exploration and exploitation, Journal of Business Research (2014), http://dx.doi.org/10.1016/j.jbusres.2014.07.011

C. Gatti et al. / Journal of Business Research xxx (2014) xxx–xxx

references — as provided by the U.S. Patent and Trademark Office for each considered patent. In addition, we refer to the concordance table between U.S. patent subclasses and the SIC codes released by USPTO at the end of our sample time frame (see Chung & Yeaple, 2008). Following Vagnani (2012) and Engelsman and van Raan (1994), for each possible combination of subclasses in a given industry, we calculate the number Ri,j,g,Δt of raw co-occurrences, that is the number of times that two subclasses belonging to a given industry have been jointly considered by inventors in a given period Δt. To make the results comparable across different industries and to control for the effect of using particularly popular combinations of subclasses (Henderson, 1995), we normalize the raw number of co-occurrences with the expected average number of co-occurrences and its standard deviation (see also Teece, Rumelt, Dosi, & Winter, 1994; Nesta & Saviotti, 2006). By averaging the normalized number of co-occurrences over the total number of subclasses of a given industry, we obtain our measure of the level of interdependence. To measure the level of decomposability of firms' activities, we instead refer to a graph, defined by jointly considering subclasses as its nodes and the associate numbers of raw co-occurrences as ties. We draw insight from the network analysis literature (Watts & Strogatz, 1998; Yayavaram & Ahuja, 2008), and distinguish subclasses that are within the same clusters from those being across clusters based on the following conditions: (1) the presence of a strong tie between them and of at least one common subclass to which both are tied; (2) the presence of a strong tie between them without any other ties at all or (3) the presence of a weak tie between them with at least one common subclass to which both are tied. In order to identify the cutoff value for classifying ties as strong or weak, we generate a null matrix by applying random permutations of the matrix associated to the graph of cooccurrences for each industry, assuming fixed column and row sums. Note that the proposed procedure, replicated one thousand times, makes our measure conservative, since it does not require any assumption on the form of the distribution of co-occurrences among different subclasses (Sanderson, Moulton, & Selfridge, 1998). Across different repetitions and for each combination of subclasses included in a given industry at time t, we develop a computer program which allows counting the number of times (denoted with Vi,j,g,Δt) that the observed co-occurrence is greater than the value that occurs in the null matrix. For each subclass included in a given SIC-based product field at time t, we evaluate its integration with neighboring subclasses outside its cluster (Integrationi,g,Δt) as the number of ties that involve out of cluster subclasses divided by the maximum possible number of ties. The latter is equal to Ji · ( Ji − 1) / 2 and Ji is the number of subclasses to which the focal subclass i is linked. We assume integration for a subclass to be zero when it has no neighbors. The level of decomposability in a given industry at time t is measured as (1 − weighted sum of the integration for each subclass), with the percentage of patents that belong to each subclass as the weight. We estimate the decomposability for each year from 1989 to 2008. It must be observed that our measures of interdependence and decomposability may vary between industries but not within a single industry. In that, they differ from previous measures considering interdependence as a specific trait of inventions (Fleming & Sorenson, 2004), or as expressing the interaction, at the intra-organizational level, between knowledge, structure and production processes (Brusoni et al., 2007; Yayavaram & Ahuja, 2008). Since interdependence and decomposability are industry-level attributes, we also account for the possibility that firms operate in multiple industries. Accordingly, from Compustat Industry Segment data and for 1989–2008, we extract the primary Standard Industrial Classification codes associated with the business segments in which our sample organizations operate (see Kahle & Walkling, 1996 for a discussion on SIC codes), and relate each business segment (via its primary SIC code) to the SIC-based product field through a specific computer program. We express the levels of interdependence and those of decomposability faced by each organization in a given year as the sales-

5

weighted average of the levels of interdependence and decomposability associated with each firm's SIC-based product field. We test the robustness of our results by using an asset-weighted level of decomposability instead of a sales-weighted level. We further use the primary SIC code associated with the segment with the highest level of sales as a reference. In both cases our results stand. 3.1.2. Exploration and exploitation measures Existing research similarly provides several ways to measure exploration and exploitation, although there are no widely accepted indicators (see, for example, Lavie et al., 2010). In particular, one might observe that researchers have mostly relied on two different methods to determine exploration and exploitation measures. On the one hand, a number of contributions has developed survey-based measures of exploration and exploitation, resulting from analyzing interview data on R&D activities and innovation processes (e.g., He & Wong, 2004; Lubatkin, Simsek, Ling, & Veiga, 2006). On the other hand, other scholars have referred to patent data, which can be easily collected, provide detailed information and allow making comparisons (see, for example, Rosenkopf & Nerkar, 2001; Katila & Ahuja, 2002). Although recognizing that existing research offers a sound basis on which one might build, we resort to content analysis to develop our measures of exploration and exploitation in order to overcome some potential limitations that affect the two mentioned approaches. A survey-based approach may indeed not be appropriate to deal with a 20 year-longitudinal study, as in our case, due to the introduction of potential biases as a consequence of endogeneity problems (Kleinknecht, Van Montfort, & Brouwer, 2002). The longitudinal nature of our study in fact requires data on intraorganization problem-solving behaviors over a long period that are not always public, are time consuming to gather and subject to errors and bias (Katila, 2002). On the flip side, despite suiting a longitudinal research design and allowing comparison at the industry-level of analysis, patent-based data also appears limited to determine firm-level measures of exploration and exploitation, particularly when the sample is composed by firms operating in different industries (Griliches, 1990). Provided that any measure should be consistent with the conceptual definitions of exploration and exploitation (Gupta, Smith, & Shalley, 2006), we thus resort to content analysis mainly because (a) it allows deriving exploration and exploitation directly from management discourses and, at the same time, makes it possible to consider such measures over an extended time period; (b) it permits to catch differences in managerial orientation across various industries, and (c) makes it easier to control for endogeneity issues that may affect estimates (see also Michalisin, 2001). Finally, through CATA, multiple texts can be analyzed without the errors or biases associated with human coders (Stevenson, 2001). We follow March's (1991: 71) original definitions of exploration — captured by terms such as “search, variation, risk taking, experimentation, play, flexibility, discovery, innovation” and of exploitation — captured by terms such as “refinement, choice, production, efficiency, selection, implementation, execution”. For each word considered, we generate all possible variants, and perform a key-word-in-context analysis (Krippendorff, 2004) to identify and eliminate inconsistencies or irrelevant expressions. In addition, we formulate a list of words (e.g., we hope to, we will, we plan to) whose occurrence signals a potential future-oriented strategic intention. A computer program is generated that marks with a special character all sentences in which inconsistencies and future-oriented strategic intentions occur. We then build our proxy measure for firms' orientation toward exploratory (exploitative) activities by counting the words included in the dictionaries for all the available documents, excluding those marked with a special character. Because some firms have longer documents and therefore the incidence of using key words in the content analysis may be stronger, we divide the total number of explorative (exploitative) words by the total number of words included in each document.

Please cite this article as: Gatti, C., et al., Interdependence among productive activities: Implications for exploration and exploitation, Journal of Business Research (2014), http://dx.doi.org/10.1016/j.jbusres.2014.07.011

6

C. Gatti et al. / Journal of Business Research xxx (2014) xxx–xxx

3.1.3. Reliability of independent variables We assess the convergent validity of our measures of exploration and exploitation by randomly selecting 60 firms per year and acquire via Factiva all Reuters news related to the selected organizations in the given years. We measure the inter-coder reliability of different coders and observe the tendency of different people to code the text the same way. We also note a significant Pearson correlation coefficient of 0.69 (0.71) (p b 0.01) between manual classification and our automated word count operationalization of exploration and exploitation. Another concern is the effective implementation of the declared orientation (March & Sutton, 1997). Thus, with the subsample of 60 firms, we correlate our measures for all available years with the measures of search depth and scope, as defined in Katila and Ahuja (2002), which directly capture the extent of firm exploitation and exploration in inventive activities. The Pearson correlation coefficients between our measures and the measures of exploration and exploitation proposed by Katila and Ahuja (2002) are equal to 0.56 and 0.54 and are both highly significant (p b 0.01). In addition, Bierly and Chakrabarti (1996) classified Abbott and Pfizer as “explorers” according to the radicalness of innovations introduced in the period 1989–1991. For these firms, our numbers of explorative (exploitative) words are 9 (3) and 25 (5), respectively. American Cynamid, Warner Lambert, and Johnson & Johnson are also considered “exploiters”. The number of explorative (exploitative) words for these firms equals 4 (7), 3 (10), and 8 (6), respectively. Since previous studies have pointed out that firms' attention toward exploration tends to be quite stable across a number of years (Bierly & Chakrabarti, 1996; He & Wong, 2004), we assess the temporal stability of our measures of exploration and exploitation by studying their autocorrelation functions. For both, we observe for a period T1–T0 a correlation coefficients of 0.4 that is comparable with the results of Kabanoff and Keegan (2007). From the proposed tests, we observe that our measures are generally reliable and can be used to assess firms' attention toward exploration and exploitation activities. We also test for the dimensionality of our two constructs. Particularly, as suggested by Short, Broberg, Cogliser, and Brigham (2010), if two or more dimensions exhibit too high of a correlation (more than .80), one may consider collapsing the two dimensions to form a single measure. In our sample, the correlation coefficient between exploration and exploitation is positive and lower than 0.5, which is consistent with Bierly and Daly (2007) conceptualization of exploration and exploitation as two distinct dimensions of a firm's search strategies. Concerning interdependence, we relate our measure to other comparable industry-level measures developed within extant literature. We refer to the work of Lenox et al. (2010). Using data from the 1994 Carnegie Mellon Survey (CMS) of Industrial R&D, Lenox et al. (2010) derive managers' perceptions of the complexity related to the introduction of new processes and new products. Such perception represents in their study an indication of the level of interdependencies at the industry level. The Pearson correlation coefficient between the simple average number of raw co-occurrences per subclasses in each industry (for the comparable year 1995) and the measure of interdependence proposed by Lenox et al. (2010) equals 0.41. We further consider the study of Zhou (2013) which uses the input–output tables to measure the decomposability of interdependencies at the industry level. Input–output tables represent the flow of goods between industries provided by the Bureau of Economic Activities (BEA). As Zhou (2013), we use the tables for 1992 and replicate our measure on the input–output data. We observe that the Pearson correlation coefficient between measures of decomposability based on patent data and on input–output data is equal to 0.39 and again is highly significant. Note that such correlation, although not perfect, presents a magnitude and significance level comparable to those found by Fleming and Sorenson (2004) when they correlated their measure of coupling based on patents with a measure they derived from surveying inventors about the coupling of the components of their inventions.

Moreover, if our measures of interdependence represent an attribute of the environment in which a firm operates, they must be characterized by considerable temporal stability. In order to verify the presence of this requisite, we calculate the auto-correlation coefficients between times [t, t − 1] and [t, t − 2]. Both for the levels of interdependence and decomposability, coefficients are greater than 0.90, which confirms the high temporal stability of our measures. We further test for potential reverse causality in our data. We refer to the Granger (1969) causality test, which has been often used in economics and other social science research as an important step toward disentangling causality from association. In particular, we fit a vector auto-regressive model (VAR) of order one for our measures of interdependence. Residuals for the model are regressed against various firm level measures, such as attention toward exploratory activities and Research and Development expenses. From the analysis we observe that in the sample data the associated coefficients with our measures of interdependence and decomposability are generally not significant, which signals that the potential risk of a reverse causality is limited in our design. 3.2. Dependent variable In this study we refer to financial performance as our dependent variable since it has been shown to directly capture organizational potential for firms' long run performance (Chakravarthy, 1986; Lubatkin & Shrieves, 1986; Richard, Devinney, Yip, & Johnson, 2009). We choose to employ a modified version of Tobin's q. This market-based measure of performance reflects ex ante financial market valuation of the rents expected to accrue to the firm, including its accounting profits (Fisher & McGowan, 1983). The q value is thus able to capture both search costs and benefits associated to exploratory and exploitative activities. It is also subject to smaller average errors than accounting-based measures (Chakravarthy, 1986; McFarland, 1988). Moreover, being an aggregate evaluation of a firm's long run performance, the q value better incorporates both prospects for future earnings (Richard et al., 2009), and the broad implications of interdependencies and search strategies for organizations. Tobin's q also depends on environmental characteristics (Lindenberg & Ross, 1981), and allows controlling for unobserved heterogeneity, in particular, making it possible to control for the scale of a firm's investments in search activities (Griliches, 1981). This study uses Chung and Pruitt's (1994) approximation to obtain the q value, which is highly correlated with Tobin's q as calculated through the more theoretically correct Lindenberg and Ross (1981) model. However, to control for possible measurement errors induced in our variable by consideration of historical (i.e., assets book value) rather than current replacement costs (Chung & Pruitt, 1994), we initially consider the ratio of the natural logarithm of firm market value (Log MV) to replacement cost of assets at the end of the fiscal year for each firm. In particular, the natural logarithm of firm market value is calculated as: log (price of outstanding common shares × number of shares + book value of preferred stock + book value of debt). Hall, Jaffe, and Trajtenberg (2001) note that the coefficient of Log A is unity under constant returns to scale or linear homogeneity of the value function. Following the authors, we move the log of total assets to the left hand side of the equation, and estimate the model with the log MV as the dependent variable. This makes our test more conservative in that potential measurement errors are confined to the independent variable (Arellano, 2003). We also run additional regression analyses (not reported here) with accounting-based measures. These robustness tests produce consistent results. 3.3. Control variables Following Griliches (1981), we control for the scale of investments in intangible assets, as approximated by R&D stock. The latter is derived from R&D expenditures under the assumption of a depreciation rate of 15% per annum. Implementing the procedure proposed by Hall (1990,

Please cite this article as: Gatti, C., et al., Interdependence among productive activities: Implications for exploration and exploitation, Journal of Business Research (2014), http://dx.doi.org/10.1016/j.jbusres.2014.07.011

C. Gatti et al. / Journal of Business Research xxx (2014) xxx–xxx

1993), our computer software accounts for missing observations. The time frame ranges from 1967 to 2008. Our results are robust and do not change even if one substitutes the R&D stock with the flow of R&D expenses. Additionally, we consider brand value as a product of advertising expenditures (ADV), taken in their yearly amount in order to overcome missing values biases (Lenox et al., 2010). Both the R&D stocks and ADV expenditures are normalized over the total value of assets (A). Note that our results are again robust to alternative assumptions, as when we omit advertising from the model entirely or assume unreported values are zero and include a dummy variable to mark missing values that have been substituted by zero. For the q value, we control for any market power or long run profitability of firms that is not specifically related to the search strategies of organizations. It must be observed that controlling for market power makes our estimates to account for the possibility that firms with market power tend to be less influenced by the environment (Miller, 1987). Accordingly, we use Compustat data to calculate firm cash flow (CF), measured as the two-year moving average of the ratio, that is earnings before interest and taxes plus depreciation minus taxes, divided by total assets. Cash flow is net of R&D and ADV expenses. We also control for future growth prospects by including the contemporaneous logarithm of the growth rate of sales (ΔLogS) (Hall, 1993). Firm size is controlled by the logarithm of total assets (Log A). Finally to avoid distortions in estimates due to changes in the composition of the panel related to the occurrence of failures in some organizations, and mergers and acquisitions in others (Baltagi, 2001), we use Compustat note 35 (i.e., reasons for deletion), and include a dummy variable that controls for possible firm exit during the period. Under the U.S. Federal Law, algorithms incorporated in a software program were not recognized as a patentable subject matter until after 1985. This means that before 1985, software companies primarily used non-patent methods to protect their innovations. We therefore test the robustness of our results against different sample specifications. Results hold if we omit software firms from the sample, simply consider software firms, or consider software firms and include a dummy variable to mark observations related to such firms. 4. Model specification We test our hypotheses under the assumption that endogeneity characterizes the way exploration and exploitation activities are determined under firm-specific and industry-specific factors. In that, we introduce a system generalized methods of moments (GMM) estimator (Blundell & Bond, 1998b) that allows controlling for distortions induced by endogeneity on estimates, and has been recently used in the management field (e.g., Cassar, 2010; Keil, Maula, Schildt, & Zahra, 2008). Thanks to our choice of a system GMM estimator, we include lagged values of the dependent variable, which further increases the robustness of estimates (Jacobson, 1990). Interaction effects are considered as an endogenous function of the exogenous variables (Billett, King, & Mauer, 2007; Greene, 2002). Following conventional GMM estimations, we introduce a three-period lag to instrumentalize independent variables. We also test for the effects on estimates related to the use of an excessive number of instruments. Therefore, we introduce a minimal number of instruments by limiting lag depth to one (Roodman, 2009). Coefficients preserve their signs and significance (see Model 6 of Table 2). It must be observed that using a system GMM estimator when the dependent variable follows a stationary process provides unbiased estimates (Blundell & Bond, 1998a). To this end, it is required that deviations from the value of the dependent variable are distributed randomly and are not correlated with any independent variables. We test for these conditions performing the Hansen J test for overidentifying restrictions, as suggested in models that include standard errors that are robust to heteroskedasticity (Hansen, 1982). We also report autocorrelation coefficients in second (z2) differences and test

7

for multicollinearity. Concerning the former, as observed by Arellano (2003: 121) “if the errors in levels are serially independent, those in first-differences will exhibit first-but not second-order serial correlation.” Concerning the latter, we perform estimates in which variables are progressively introduced. In addition, for each model we report the mean variance inflation factor (VIF), which is an indicator of the severity of multicollinearity among independent variables. Finally, as proposed by Cohen (1978), in testing our hypotheses we introduce a hierarchical step-up approach; that is, the main effect is tested in an equation containing only that effect and interactions are then considered for their contribution over and above main effect. We then report the Wald chi-squared test of the marginal contribution of the added variables to model fit. Finally, consistently with the use of a system GMM estimator, the overall model significance is assessed by the Wald χ2 test (Arellano, 2003). 5. Estimates and results Table 1 provides descriptive statistics. The limited magnitude of the correlation coefficient between interdependence and decomposability indicates that these variables are distinct industry attributes. As suggested by Aiken and West (1991), the variables exploration, exploitation, interdependence, and decomposability and Log A are mean centered to minimize the potential for multicollinearity in equations for interaction terms. Table 2 depicts the panel regression results for Log MV. Model 1 reports the regression with only the control variables. Model 2 introduces the main effects of exploration, exploitation, interdependence and decomposability. Particularly, exploration offers a positive contribution to firms' long-run performance, while the effect of exploitation is negative. H1 proposes that the level of interdependence among firms' activities moderates the effects of exploitation and exploration to firms' long-run financial performance. Model 3 supports H1. The interaction term between exploration (exploitation) and interdependence is positive (negative) and significant. Therefore, greater levels of exploration (exploitation) activities contribute more positively (negatively) to long-run financial performance of firms that operate in industries that exhibit extended level of interdependence. Note that, compared with the model with only the main effects (see Model 2), the marginal Wald chi-squared statistics for Model 3 is equal to 6.61 and such coefficient is significant (p b 0.05, df = 2). H2 suggests that the higher the level of decomposability, the lower (greater) the contribution of exploration (exploitation) to firms' longrun financial performance. Model 4 provides evidence that supports H2. In particular, the interaction term between exploration (exploitation) and decomposability is negative (positive) and statistically significant. Therefore, our analysis suggests that in firms that face high levels of decomposability of productive activities at the industry level, a greater attention to exploration (exploitation) contributes less (more) to firms' long-run financial performance. Compared with Model 2 of Table 2, the marginal Wald chi-squared statistic for Model 4 is equal to 14.02 and such coefficient is again statistically significant (p b 0.001, df = 2). Finally, our results hold in models either where interactions effects are jointly considered in the same equation (Model 5). We also test for possible biases induced by instrument proliferation (Roodman, 2009). In particular, our results hold even if we consider a one-period lag for instrumentalizing independent variables (see Model 6). To better appreciate our results we graph the interaction effects. Exploration, exploitation, interdependence, and decomposability take the values of one standard deviation above (i.e., high level) and one standard deviation below (i.e., low level) the mean, respectively. All other variables are assumed at their mean value. Fig. 1 shows that exploratory (exploitative) activities offer a greater positive contribution to long-run performance when firms face either more extended (more

Please cite this article as: Gatti, C., et al., Interdependence among productive activities: Implications for exploration and exploitation, Journal of Business Research (2014), http://dx.doi.org/10.1016/j.jbusres.2014.07.011

1 −0.060⁎ 1 −0.732⁎ −0.222⁎ 1 −0.077⁎ −0.006 0.033 1 0.044⁎ 0.272⁎ −0.125⁎ 0.010 1 −0.019 0.003 0.021 0.028 0.082⁎ 1 −0.198⁎ −0.120⁎ 0.155⁎ 0.003 −0.037 0.445⁎ 1 −0.212⁎ 0.1113⁎ 0.1186⁎ −0.0226 0.1087⁎ 0.1781⁎ 0.3290⁎ 1

limited) levels of interdependence among their activities or a more limited (greater) level of decomposability of such activities. All models include periodic effects that are omitted from Table 2 but which are generally significant. Variables representing the stock of intangible assets, particularly those related to R&D activities, are significant and contribute to market value, as already observed in the literature (Griliches, 1981). Our estimates reveal the presence of heteroskedasticity. For example, considering model 1 of Table 2, and running a simple OLS regression, the White–Koenker test statistic has a χ2(30) of 417.96 (p b 0.001). Thus, in all estimates we consider standard errors that are robust to heteroskedasticity. In addition, we observe the presence of endogeneity in our estimates but their effects on the latter are well controlled by our system GMM estimator. All models show non-significant second-order difference autocorrelation coefficients z2. Hypotheses of the exogeneity of instruments are not rejected by the Hansen J test. Note that for all models VIF values are far below the rule-of-thumb cutoff of 10 (Neter, Wasserman, & Kutner, 1985). 6. Discussion and conclusions

N = 7026; robust standard errors. The sample period is 1989–2008. Exploration and exploitation are expressed as value × 103. ⁎ p b .05.

1 0.330⁎ 1 0.028 0.076⁎ 1 ⁎ 0.100 0.129⁎ 0.901⁎ 1 ⁎ −0.207 −0.263⁎ −0.082⁎ −0.168⁎ ⁎ −0.117 0.153⁎ 0.069⁎ −0.008 0.159⁎ 0.137⁎ 0.137⁎ 0.138⁎ 0.001 0.000 0.047⁎ −0.034 −0.041 0.147⁎ 0.247⁎ 0.136⁎ 0.436⁎ 0.214⁎ 0.231⁎ 0.219⁎ ⁎ 0.850 0.398⁎ 0.022 0.096⁎ 0.339⁎ 0.685⁎ 0.087⁎ 0.140⁎ 1 0.444⁎ 0.181⁎ 0.230⁎ 0.215⁎ −0.017 0.006 0.022 0.030 0.087⁎ 0.749⁎ 0.420⁎ 0.188⁎ 1 0.084⁎ −0.038 0.110⁎ 0.246⁎ 0.134⁎ 0.050⁎ 0.273⁎ −0.131⁎ 0.014 0.989⁎ 0.083⁎ −0.040 0.104⁎ 1 0.011 0.030 0.008 −0.013 −0.006 −0.076⁎ 0.053⁎ −0.019 −0.053⁎ 0.078⁎ 0.018 0.027 −0.001 −0.018 1 0.044⁎ −0.124⁎ 0.027 0.157⁎ 0.120⁎ 0.142⁎ 0.138⁎ −0.690⁎ −0.213⁎ 0.913⁎ 0.095⁎ −0.114⁎ 0.031 0.151⁎ 0.119⁎ 1 −0.221⁎ −0.007 0.267⁎ 0.002 −0.118⁎ 0.113⁎ 0.072⁎ −0.003 −0.065⁎ 0.952⁎ −0.201⁎ −0.003 0.264⁎ 0.004 −0.118⁎ 0.110⁎ 1 −0.060⁎ −0.727⁎ −0.085⁎ 0.047⁎ −0.022 −0.201⁎ −0.210⁎ −0.079⁎ −0.156⁎ 0.934⁎ −0.059⁎ −0.695⁎ −0.064⁎ 0.040 −0.026 −0.198⁎ −0.207⁎ 1 −0.169⁎ −0.011 0.145⁎ −0.024 0.139⁎ 0.220⁎ 0.099⁎ 0.141⁎ 0.906⁎ 0.989⁎ −0.163⁎ −0.007 0.144⁎ −0.006 0.139⁎ 0.223⁎ 0.095⁎ 0.137⁎

19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1

1 0.902⁎ −0.085⁎ 0.067⁎ 0.144⁎ 0.055⁎ 0.247⁎ 0.232⁎ 0.027 0.091⁎ 0.970⁎ 0.878⁎ −0.071⁎ 0.067⁎ 0.130⁎ 0.083⁎ 0.245⁎ 0.230⁎ 0.021 0.097⁎ 0.734 0.661 0.312 0.040 0.110 0.094 0.399 0.708 1.955 0.105 0.731 0.658 0.305 0.041 0.111 0.094 0.395 0.001 0.002 0.105

SD Mean

Table 1 Descriptive statistics and pairwise correlations for sampled firms.

3.447 3.382 0.194 0.017 0.065 0.029 0.320 0.591 2.455 0.729 3.443 3.374 0.192 0.017 0.065 0.030 0.316 0.001 0.002 0.730

20

C. Gatti et al. / Journal of Business Research xxx (2014) xxx–xxx

1. Log MV 2. Log A 3. R&D/A 4. ADV/A 5. CF/A 6. ΔlogS 7. Interdependence 8. Exploration 9. Exploitation 10. Decomposability 11. Log MVt − 1 12. Log At − 1 13. R&D/At − 1 14. ADV/At − 1 15. CF/At − 1 16. ΔlogSt − 1 17. Interdependencet − 1 18. Exploitationt − 1 19. Explorationt − 1 20. Decomposabilityt − 1

8

This paper elaborates on the interdependence construct and discusses its implications on the net benefits of exploration and exploitation activities. We formulate reliable measures of interdependence based on public available data, consider a longitudinal research design to better account for the long-term implications of exploration and exploitation activities, and introduce a system GMM estimator to control for potential sources of endogeneity in estimates. We have tested and found that greater levels of exploration contribute more to long-run financial performance of firms that face more extended interdependencies and low levels of decomposability. On the other hand, greater levels of exploitation contribute more to long-run financial performance of firms that face more limited interdependencies and high levels of decomposability. Interdependence thus matters and has a relevant impact on the relationship between exploration, exploitation and firms' longrun financial performance in much the same way as other contingent factors, which have been extensively analyzed in literature, such as organizational size, environmental dynamism, competitiveness and munificence (Lavie et al., 2010). The study presents the levels of interdependencies among firms' productive activities and the distribution patterns of those interdependencies as distinct attributes of the environment in which organizations operate and shows that both influence the long-term performance implications of exploration and exploitation activities. In that, our findings further understanding of the conditions under which greater levels of exploration and exploitation are necessary for a firm to sustain its long-run performance (Gupta et al., 2006). Moreover, our analysis provides empirical ground to theoretical studies dealing with exploration, exploitation, interdependence, decomposability, and firms' long-term performance. Considering that there are some computer-based simulation studies illustrating how greater levels of decomposability make exploration more essential and exploitation less essential to long-run organizational performance (Rivkin & Siggelkow, 2007), one may argue if this is the case here as well. Our data indeed shows exploration to be less essential while exploitation more essential in industries that exhibit greater levels of decomposability. Perhaps, even when the levels of interdependence are relatively high, a high level of decomposability corresponds to lower rather than greater complexity faced by firms in their search activities, which makes exploitation more beneficial to firms' long-run performance. 6.1. Limitations and future research We believe that our work has only started to scratch the surface of a thorough understanding of the role of interdependence among productive activities, since it presents some limitations, which however pave

Please cite this article as: Gatti, C., et al., Interdependence among productive activities: Implications for exploration and exploitation, Journal of Business Research (2014), http://dx.doi.org/10.1016/j.jbusres.2014.07.011

C. Gatti et al. / Journal of Business Research xxx (2014) xxx–xxx

9

Table 2 Results of system GMM estimator for log of market value. Dependent: Log MV

Model 1

Model 2

Model 3

Model 4

Model 5

Model 6

Intercept Log A R&D/A ADV/A CF/A ΔlogS Exploitation Exploration Interdependence Decomposability Exploitation × Interdependence Exploration × Interdependence Exploitation × Decomposability Exploration × Decomposability Log MVt − 1 z2 Observations Wald χ2 (df) No. of instruments Hansen J test Mean VIF

0.555⁎⁎⁎ (0.077) 0.188⁎⁎⁎ (0.030) 0.120⁎⁎⁎ (0.030) 0.426⁎⁎ (0.140) 0.308⁎⁎ (0.092) 0.660⁎⁎⁎ (0.139)

0.646⁎⁎⁎ (0.079) 0.222⁎⁎⁎ (0.027) 0.099⁎⁎⁎ (0.028) 0.275⁎(0.136) 0.241⁎⁎ (0.090) 0.663⁎⁎⁎ (0.147) −11.75⁎⁎ (3.728) 40.74⁎⁎ (11.73) 0.048⁎⁎⁎ (0.010) −0.082 (0.054)

0.645⁎⁎⁎ (0.085) 0.222⁎⁎⁎ (0.027) 0.104⁎⁎⁎ (0.028) 0.281⁎(0.131) 0.256⁎⁎ (0.087) 0.660⁎⁎⁎ (0.150) −8.721⁎⁎ (3.106) 28.85⁎⁎ (8.668) 0.043⁎⁎⁎ (0.010) −0.086† (0.051) −12.97⁎ (6.342) 60.54⁎⁎ (25.13)

0.654⁎⁎⁎ (0.074) 0.226⁎⁎⁎ (0.025) 0.098⁎⁎ (0.028) 0.287⁎(0.128) 0.246⁎⁎ (0.091) 0.657⁎⁎⁎ (0.148) −11.67⁎⁎ (4.056) 42.23⁎⁎ (11.67) 0.048⁎⁎⁎ (0.009) −0.072 (0.050)

0.649⁎⁎⁎ (0.083) 0.223⁎⁎⁎ (0.026) 0.104⁎⁎⁎ (0.028) 0.294⁎ (0.122) 0.265⁎⁎ (0.089) 0.654⁎⁎⁎ (0.148) −9.863⁎⁎ (3.602) 29.23⁎⁎ (10.48) 0.042⁎⁎⁎ (0.009) −0.055 (0.046) −11.50⁎ (5.703) 63.41⁎⁎ (21.96) 67.80⁎⁎ (26.15) −199.5⁎⁎ (67.54) 0.773⁎⁎⁎ (0.025) 0.49 6473 41,672 (38) 1003 436.83 3.14

2.212⁎⁎⁎ (0.116) 0.665⁎⁎⁎ (0.033) 0.328⁎⁎⁎ (0.087) 0.961⁎⁎⁎ (0.260) 0.659⁎⁎⁎ (0.184) 0.734⁎⁎⁎ (0.217) −16.902⁎⁎⁎ (4.196) 47.426⁎⁎⁎ (9.060) 0.123⁎⁎⁎ (0.022) −0.150 (0.099) −20.62⁎ (8.969) 98.13⁎⁎⁎ (23.23) 145.25⁎⁎⁎ (32.60) −383.4⁎⁎⁎ (72.19) 0.275⁎⁎⁎ (0.028) 1.24 6473 8034 (38) 541 431.63 3.14

0.801⁎⁎⁎ (0.023) 0.45 6473 37,954 (30) 445 433.48 3.06

0.776⁎⁎⁎ (0.023) 0.52 6473 42,208 (34) 727 434.66 3.24

0.775⁎⁎⁎ (0.025) 0.52 6473 32,725 (36) 865 437.02 3.12

71.94⁎⁎ (28.91) −256.4⁎⁎⁎ (76.57) 0.773⁎⁎⁎ (0.022) 0.48 6473 44,642 (36) 865 439.23 3.24

Note: All estimates include year dummies. Dummies controlling for reason for deletion (i.e., Compustat note 35) also are included in each model. For the GMM all variable are considered endogenous but CF/A, ΔlogS, Interdependence and Decomposability are predetermined. Dummies variables are exogenous. For the equation in levels, ΔXt − 1,…, ΔXt − l and year dummies are used as instruments. For the equation in first differences, Xt − 1,…, Xt − l and year dummies are used as instruments. Panel unbalanced. Sample period is 1989–2008. † p b .1. ⁎ p b .05. ⁎⁎ p b .01. ⁎⁎⁎ p b .001.

the way to future research. First, it is limited to one type of interdependence. It might therefore be interesting to test the hypotheses in this paper even in light of the role of other specific types of interdependencies (e.g., interdependencies stemming from institutional, competitive and social interactions) or at different organizational levels. In this vein, an interesting research question would concern whether

interdependencies that arise from different sources add linearly in determining the complexity of the environment or whether such additional complexity is both qualitatively and quantitatively different from that stemming from higher levels of interdependence of just one kind. Second, in analyzing the moderating role of the structure of interdependencies on the exploration–exploitation–long-run performance

a) Log MV

Log MV

b)

Low

Exploitation

Low Interdependence

Low Interdependence

High Interdependence

High Interdependence

High

Low

High

Exploration

d)

Log MV

Log MV

c)

Low

Low Decomposability

Low Decomposability

High Decomposability

High Decomposability

Exploitation

High

Low

Exploration

High

Note: Curves are generated from Model 5 in Table 2. Coefficients that are not significant are omitted from equations.

Fig. 1. The long-run financial performance of exploration and exploitation activities. Note: Curves are generated from Model 5 in Table 2. Coefficients that are not significant are omitted from equations.

Please cite this article as: Gatti, C., et al., Interdependence among productive activities: Implications for exploration and exploitation, Journal of Business Research (2014), http://dx.doi.org/10.1016/j.jbusres.2014.07.011

10

C. Gatti et al. / Journal of Business Research xxx (2014) xxx–xxx

relationships, we have referred to broad definitions of such constructs. Additional insight would come from considering more specific acts of exploration and exploitation (e.g., new generations of products, refinement of existing products, development of new process technologies) or acts of exploration and exploitation that stem from firms collective efforts (e.g., participation in standard setting organizations, R&D consortium) and test their effects on more specific measures of performance (such as the number of new products and total sales accounted for by new products). As an example, by participating in a standard setting organization, firms may willingly agree to a more limited or even suboptimal solution to a technological problem in exchange for the ability to free up resources for other product activities. Alternatively, by engaging in R&D consortia, firms may be able to conduct research into otherwise inaccessible technological areas and expand exploration. Third, interdependence has been conceived as a moderating variable between exploratory–exploitative activities and firms' long-run financial performance. One may enrich the framework and relate interdependencies to different but relevant aspects of firms' competitive strategies, including imitation (Rivkin, 2000) and replication of success (Rivkin, 2001). Moreover, our analysis could be further extended to consider the ambidexterity construct, i.e. firms' ability to achieve both exploration and exploitation orientations (e.g., He & Wong, 2004; Hess & Rothaermel, 2009), and test whether such a pursuit results in higher performance depending on the levels of interdependence and decomposability. According to March (1991), firms indeed face inherent difficulties in attaining and maintaining a proper balance between exploration and exploitation, running the risk of being mediocre at both. Recent contributions have only started to explore interdependence between a firm's exploration and exploitation orientations, for instance with respect to their influence on strategic learning (Sirén, Kohtamäki, & Kuckertz, 2012). The real challenge for future research would be to discern whether and how firms may benefit from combining exploration and exploitation and overcoming inherent trade-offs that characterize these search activities, given the level and the structure of interdependencies that characterize the industry in which a firm operates. At the top of our agenda lies however the possibility to directly model how a focal firm can challenge the potential for interdependence faced by all firms in a given industry, and how changes in such potential may reverberate on the focal firm's long-run financial performance. Consider the example of the computing industry. In this industry, computing platforms or architectural standards, rather than firms, are the primary mechanisms for competitive entry and main drivers for interdependencies in the industry as a whole. Platforms and standards exist among firms within the same industry and are driven by meta-organizations, such as standard setting organizations and R&D consortia. As Rysman and Simcoe (2008: 1920), explain “voluntary standard-setting organizations are a common feature of systems industries, where firms supply interoperable components for a shared technology platform. These institutions promote coordinated innovation by providing a forum for collective decision making and a potential solution to the problem of fragmented and overlapping intellectual property rights”. Firms have managerial discretion whether or not to participate in these meta-organizations. One can then theorize how participation in the latter may enable a firm to expand its exploration activities, favor the introduction of new combinations of productive activities and, in doing so, alter the structure of interdependencies and shape the potential for interdependence faced by all firms in a given industry in a direction that is likely to sustain the focal firm's long-run performance (Bresnahan & Greenstein, 1999). 6.2. Implications for practice Our study also has implications for managerial practice. In particular, managers that cannot or do not see the opportunity to act on the environment should understand that, in industries where interdependence

among firms' activities is limited whereas decomposability is high, firms can embrace search strategies that are more local and more directed toward improvement within specific clusters of productive activities, and thus increase their long-run performance. Consider the following example: the house appliance industry shows a high level of decomposability of productive activities. Accordingly, in this industry, high and lowexploration firms tend to show reduced differences in terms of financial performance. Specifically, in the period 1989–1995, Whirlpool posed a more limited attention toward broad exploration of new possibilities. Beyond this, the firm developed basic platforms, common components, technologies and manufacturing processes, exploiting these common components and combining them in different ways. As a consequence, Whirlpool was able to introduce many new products to the market, and to reduce the gap with other more exploratory organizations (Maruca, 1994). Similar patterns can be found in other industries with high levels of decomposability, such as motorcycles, bicycles and parts, guided missiles and space vehicles, motor vehicles and other motor vehicle equipment, and railroad equipment. Conversely, for firms that face high levels of interdependencies and low levels of decomposability, pursuing broader exploration remains highly beneficial and necessary for long-run performance. The plastics materials and synthetic resins industry offers an example of an environment with a low level of decomposability. In this industry, the attention toward broad exploration is an important driver of firms' superior performance (e.g., see Arora, Landau, & Rosenberg, 1999). In fact, high-exploration firms, such as DuPont, Dow Chemical, and Rohm and Haas, tend to show significantly greater financial performance than low-exploration firms (e.g., Wellman, Schulman, Dexter). Such pattern is also observable in comparable industries, such as agricultural chemicals, petroleum and natural gas extraction and refining, drugs and medicines. Managers are then advised to either find sufficient resources to be devoted to broad exploration activities or to evaluate the opportunity of divesting from such industries, selling related activities to firms that could benefit from them. In addition, managers are also required to set forth routines, incentives, and structures to maintain greater focus on exploration despite natural tendencies to limit these activities. However, managers can do more. They can act to challenge the potential of interdependence among firms' activities in a direction that enlarges the net benefits of firms' exploratory and exploitative activities. The case of Shimano is exemplary: the investments made in the mid1980s allowed the firm to introduce different breakthrough innovations (in particular the index shifting system). Given the strong cause–effect relationship from firm's action, via its product design choices, to the structure of interdependencies that characterized the bicycle industry, these breakthrough innovations had the further consequence of reducing the level of decomposability in the industry, which greatly reinforced the long-term benefits of the firm's exploration activities (Fixson & Park, 2008; Galvin & Morkel, 2001; Isely & Roelofs, 2004). Acknowledgments The authors are grateful to Michele Simoni for his help and support in preparing the manuscript and to the journal’s refererees for their comments on an earlier draft. The authors would also like to acknowledge financial support of the study from Sapienza University and particularly from a research grant by Regione Lazio. References Abernathy, W. J., & Clark, K. B. (1985). Mapping the winds of creative destruction. Research Policy, 14, 3–22. Abrahamson, E., & Amir, E. (1996). The association between the information contained in the president's letter to shareholders and accounting market variables. Journal of Business Finance and Accounting, 23, 1157–1182. Aiken, L. S., & West, S. G. (1991). Multiple regression: Testing and interpreting interactions. Sage: Newbury Park, CA. Arellano, M. (2003). Panel data econometrics. New York: Oxford University Press.

Please cite this article as: Gatti, C., et al., Interdependence among productive activities: Implications for exploration and exploitation, Journal of Business Research (2014), http://dx.doi.org/10.1016/j.jbusres.2014.07.011

C. Gatti et al. / Journal of Business Research xxx (2014) xxx–xxx Arora, A., Landau, R., & Rosenberg, N. (1999). Dynamics of comparative advantage in the chemical industry. In D. C. Mowery, & R. R. Nelson (Eds.), Sources of industrial leadership: Studies of seven industries (pp. 217–228). Cambridge, UK: Cambridge University Press. Baltagi, B. H. (2001). Econometric analysis of panel data. New York: John Wiley & Sons. Baumann, O., & Siggelkow, N. (2012). Dealing with complexity: Integrated vs. chunky search processes. Organization Science, 1–40 (Published online before print March 15, 2012). Benner, M. J., & Tushman, M. (2002). Process management and technological innovation: A longitudinal study of the photography and paint industries. Administrative Science Quarterly, 47, 676–706. Bierly, P., & Chakrabarti, A. (1996). Generic knowledge strategies in the U.S. pharmaceutical industry. Strategic Management Journal, 17, 123–135. Bierly, P., & Daly, P.S. (2007). Alternative knowledge strategies, competitive environment, and organizational performance in small manufacturing firms. Entrepreneurship: Theory and Practice, 31, 493–516. Billett, M. T., King, T. -H. D., & Mauer, D. C. (2007). Growth opportunities and the choice of leverage, debt maturity, and covenants. Journal of Finance, 62, 697–730. Blundell, R. W., & Bond, S. R. (1998a). Initial conditions and moment restrictions in dynamic panel data models. Journal of Econometrics, 87, 115–143. Blundell, R. W., & Bond, S. R. (1998b). GMM estimation with persistent panel data: An application to production functions. Bowman, E. H. (1984). Content analysis of annual reports for corporate strategy and risk. Interfaces, 14, 61–71. Bresnahan, T. F., & Greenstein, S. (1999). Technological competition and the structure of the computer industry. Journal of Industrial Economics, 47, 1–40. Brusoni, S., Marengo, L., Prencipe, A., & Valente, M. (2007). The value and costs of modularity: A problem-solving perspective. European Management Review, 4, 121–132. Buenstorf, G. (2005). Sequential production, modularity and technological change. Structural Change and Economic Dynamics, 16, 221–241. Cassar, G. (2010). Are individuals entering self-employment overly optimistic? An empirical test of plans and projections on nascent entrepreneur expectations. Strategic Entrepreneurship Journal, 31, 822–840. Cassiman, B., & Veugelers, R. (2006). In search of complementarity in innovation strategy: Internal R&D and external knowledge acquisition. Management Science, 52, 68–82. Chakravarthy, B. (1986). Measuring strategic performance. Strategic Management Journal, 7, 437–458. Chen, W. -R. (2008). Determinants of firm's backward- and forward-looking R&D search behavior. Organization Science, 19, 609–622. Chung, K. H., & Pruitt, S. W. (1994). A simple approximation of Tobin's q. Financial Management, 23, 70–74. Chung, W., & Yeaple, S. (2008). International knowledge sourcing: Evidence from U.S. firms expanding abroad. Strategic Management Journal, 29, 1207–1224. Cohen, J. (1978). Partialed products are interactions; partialed vectors are curve components. Psychological Bulletin, 85, 858–866. Courtis, J. K. (1982). Private shareholder response to corporate annual report. Accounting and Finance, 22, 53–72. Danneels, E. (2002). The dynamics of product innovation and firm competences. Strategic Management Journal, 23, 1095–1121. Davis, J., Eisenhardt, K., & Bingham, C. (2007). Developing theory through simulation methods. Academy of Management Review, 32, 480–499. Engelsman, E. C., & van Raan, A. F. J. (1994). A patent-based cartography of technology. Research Policy, 23, 1–26. Ethiraj, S. K. (2007). Allocation of inventive effort in complex product systems. Strategic Management Journal, 28, 563–584. Ethiraj, S. K., & Levinthal, D. (2004). Modularity and innovation in complex systems. Management Science, 50, 159–173. Ethiraj, S. K., Levinthal, D., & Roy, R. R. (2008). The dual role of modularity: Innovation and imitation. Management Science, 54, 939–955. Fisher, F. M., & McGowan, J. J. (1983). On the misuse of accounting rates of return to infer monopoly profits. American Economic Review, 73, 82–97. Fixson, S. K., & Park, J. -K. (2008). The power of integrality: Linkages between product architecture, innovation, and industry structure. Research Policy, 37, 1296–1316. Fleming, L. (2001). Recombinant uncertainty in technological search. Management Science, 47, 117–132. Fleming, L., & Sorenson, O. (2001). Technology as a complex adaptive system: Evidence from patent data. Research Policy, 30, 1019–1039. Fleming, L., & Sorenson, O. (2004). Science as a map in technological search. Strategic Management Journal, 25, 909–928. Frenken, K. (2006). Innovation, evolution and complexity theory. Northampton, MA: Edward Elgar. Frenken, K., Marengo, L., & Valente, M. (1999). In T. Brenner (Ed.), Interdependencies, nearly-decomposability and adaptation. Computational techniques for modelling learning in economics, Vol. 11. (pp. 145–165). Dordrecht: Kluwer. Galvin, P., & Morkel, A. (2001). The effect of product modularity on industry structure: The case of the world bicycle industry. Industry and Innovation, 8, 31–47. Ghemawat, P., & Levinthal, D. (2008). Choice interactions and business strategy. Management Science, 54, 1638–1651. Granger, C. W. J. (1969). Investigating causal relations by econometric models and crossspectral methods. Econometrica, 37, 424–438. Greene, W. H. (2002). Econometric analysis. Englewood Cliffs, NJ: Prentice Hall. Griliches, Z. (1981). Market value, R&D, and patents. Economics Letters, 7, 183–187. Griliches, Z. (1990). Patent statistics as economic indicators: A survey. Journal of Economic Literature, 28, 1661–1707. Gupta, A. K., Smith, K. G., & Shalley, C. E. (2006). The interplay between exploration and exploitation. Academy of Management Journal, 49, 693–706.

11

Hall, B. H. (1990). The manufacturing sector master file: 1959–1987. NBER working paper series, National Bureau of Economic Research. Hall, B. H. (1993). The stock market's valuation of R&D investment during the 1980s. American Economic Review, 83, 259–264. Hall, B. H., Jaffe, A. B., & Trajtenberg, M. (2001). The NBER patent citation data file: Lessons, insights and methodological tools. National Bureau of Economic Research. Hamilton, B. H., & Nickerson, J. A. (2003). Correction for endogeneity in strategic management research. Strategic Organization, 1, 51–78. Hansen, L. (1982). Large sample properties of generalized method of moment estimators. Econometrica, 50, 1029–1054. He, Z., & Wong, P. (2004). Exploration vs. exploitation: An empirical test of the ambidexterity hypothesis. Organization Science, 15, 481–494. Henderson, R. (1995). Of life cycles real and imaginary: The unexpectedly long old age of optical lithography. Research Policy, 24, 631–643. Hess, A. M., & Rothaermel, F. T. (2009). Ambidexterity and innovative performance. The role of intellectual human capital and strategic alliances, working paper. Georgia Tech University. Iansiti, M., & MacCormack, A. (1996). Team New Zealand (C). HBS Case 697–042. Ichniowski, C., & Shaw, K. (1999). The effects of human resource management systems on economic performance: An international comparison of U.S. and Japanese plants. Management Science, 45, 704–721. Isely, P., & Roelofs, M. R. (2004). Primary market and aftermarket competition in the bicycle component industry. Applied Economics, 36, 2097–2102. Jacobson, R. (1990). Unobservable effects and business performance. Marketing Science, 9, 74–85. Jansen, J. J. P., Van den Bosch, F. A. J., & Volberda, H. W. (2006). Exploratory innovation, exploitative innovation, and performance: Effects of organizational antecedents and environmental moderators. Management Science, 52, 1661–1674. Kabanoff, B., & Keegan, J. (2007). Studying strategic cognition by content analysis of annual reports: A validation involving firm innovation. Managing our intellectual and social capital: 21st ANZAM 2007 Conference (pp. 1–14). Kahle, K. M., & Walkling, R. A. (1996). The impact of industry classifications on financial research. Journal of Financial and Quantitative Analysis, 31, 309–335. Katila, R. (2002). New product search over time: Past ideas in their prime? Academy of Management Journal, 45, 995–1010. Katila, R., & Ahuja, G. (2002). Something old, something new: A longitudinal study of search behavior and new product introduction. Academy of Management Journal, 45, 1183–1194. Kauffman, S. A. (1993). The origins of order: Self-organization and selection in evolution. New York: Oxford University Press. Keil, T., Maula, M., Schildt, H., & Zahra, S. A. (2008). The effect of governance modes and relatedness of external business development activities on innovative performance. Strategic Entrepreneurship Journal, 29, 895–907. Kleinknecht, A., Van Montfort, K., & Brouwer, E. (2002). The non-trivial choice between innovation indicators. Economics of Innovation and New Technology, 11, 109–121. Krippendorff, K. (2004). Content analysis: An introduction to its methodology. Thousand Oaks, CA: Sage. Langlois, R. N. (2002). Modularity in technology and organization. Journal of Economic Behavior and Organization, 49, 19–37. Lavie, D., Stettner, U., & Tushman, M. L. (2010). Exploration and exploitation within and across organizations. Academy of Management Annals, 4, 109–155. Lee, D.M. S., & Allen, T. J. (1982). Integrating new technical staff: Implication for acquiring new technology. Management Science, 28, 1405–1420. Lenox, M. J., Rockart, S. F., & Lewin, A. Y. (2006). Interdependency, competition, and the distribution of firm and industry profits. Management Science, 52, 757–772. Lenox, M. J., Rockart, S. F., & Lewin, A. Y. (2007). Interdependency, competition, and industry dynamics. Management Science, 53, 599–615. Lenox, M. J., Rockart, S. F., & Lewin, A. Y. (2010). Does interdependency affect firm and industry profitability? An empirical test. Strategic Management Journal, 31, 121–139. Leonard-Barton, D. (1992). Core capabilities and core rigidities: A paradox in managing new product development. Strategic Management Journal, 13, 111–126. Levinthal, D. (1997). Adaptation on a rugged landscape. Management Science, 43, 934–950. Levinthal, D., & March, J. G. (1981). A model of adaptive organizational search. Journal of Economic Behavior and Organization, 2, 307–333. Levinthal, D., & March, J. G. (1993). The myopia of learning. Strategic Management Journal, 14, 95–112. Levitt, B., & March, J. G. (1988). Organizational learning. Annual Review of Sociology, 14, 319–340. Lewin, A. Y., Long, C. P., & Caroll, T. N. (1999). The coevolution of new organizational forms. Organization Science, 10, 535–550. Lindenberg, E. B., & Ross, S. A. (1981). Tobin's q ratio and industrial organization. Journal of Business, 54, 1–32. Lubatkin, M., & Shrieves, R. E. (1986). Toward reconciliation of market performance measures to strategic management research. Academy of Management Review, 11, 497–512. Lubatkin, M. H., Simsek, Z., Ling, Y., & Veiga, J. F. (2006). Ambidexterity and performance in small-to medium-sized firms: The pivotal role of top management team behavioral integration. Journal of Management, 32, 646–672. March, J. G. (1991). Exploration and exploitation in organizational learning. Organization Science, 2, 71–87. March, J. G. (2006). Rationality, foolishness, and adaptive intelligence. Strategic Entrepreneurship Journal, 27, 201–214. March, J. G. (2008). Explorations in organizations. Stanford, CA: Stanford University Press. March, J. G., & Sutton, R. I. (1997). Organizational performance as dependent variable. Organization Science, 8, 698–706.

Please cite this article as: Gatti, C., et al., Interdependence among productive activities: Implications for exploration and exploitation, Journal of Business Research (2014), http://dx.doi.org/10.1016/j.jbusres.2014.07.011

12

C. Gatti et al. / Journal of Business Research xxx (2014) xxx–xxx

Maruca, R. F. (1994). The right way to go global: An interview with Whirlpool CEO David Whitwam. Harvard Business Review, 72(2), 135–146. McFarland, H. (1988). Evaluating q as an alternative to the rate of return in measuring profitability. Review of Economics & Statistics, 70, 614–622. Michalisin, M.D. (2001). Validity of annual report assertions about innovativeness: An empirical investigation. Journal of Business Research, 53, 151–161. Miller, D. (1987). The genesis of configuration. Academy of Management Review, 12, 686–701. Nesta, L., & Saviotti, P. (2006). Firm knowledge and market value in biotechnology. Industrial and Corporate Change, 15, 625–652. Neter, J., Wasserman, W., & Kutner, M. H. (1985). Applied linear statistical models: Regression analysis of variance, and experimental design. Homewood, IL: Richard D. Irwin, Inc. Osborne, J.D., Stubbart, C. I., & Ramaprasad, A. (2001). Strategic groups and competitive enactment: A study of dynamic relationships between mental models and performance. Strategic Entrepreneurship Journal, 22, 435–454. Patel, P., & Pavitt, K. (1997). The technological competencies of the world's largest firms: Complex and path-dependent, but not much variety. Research Policy, 26, 141–156. Previts, G., Bricker, R., Robinson, T., & Young, S. (1994). A content analysis of sell-side financial analyst company reports. Accounting Horizon, 8, 55–70. Richard, P., Devinney, T., Yip, G., & Johnson, G. (2009). Measuring organizational performance: Towards methodological best practice. Journal of Management, 35, 718–804. Rivkin, J. W. (2000). Imitation of complex strategies. Management Science, 46, 824–844. Rivkin, J. W. (2001). Reproducing knowledge: Replication without imitation at moderate complexity. Organization Science, 12, 274–293. Rivkin, J. W., & Siggelkow, N. (2006). Organizing to strategize in the face of interactions: Preventing premature lock-in. Long Range Planning, 39, 591–614. Rivkin, J. W., & Siggelkow, N. (2007). Patterned interactions in complex systems: Implications for exploration. Management Science, 53, 1068–1085. Roodman, D. (2009). A note on the theme of too many instruments. Oxford Bulletin of Economics and Statistics, 71, 135–158. Rosenkopf, L., & Nerkar, A. (2001). Beyond local search: Boundary-spanning, exploration, and impact in the optical disc industry. Strategic Management Journal, 22, 287–306. Rysman, M., & Simcoe, T. (2008). Patents and the performance of voluntary standardsetting organizations. Management Science, 54, 1920–1934. Sanderson, J. G., Moulton, M. P., & Selfridge, R. G. (1998). Null matrices and the analysis of species co-occurrences. Oecologia, 116, 275–283. Shane, S., & Venkataraman, S. (2000). The promise of entrepreneurship as a field of research. Academy of Management Review, 25, 217–226. Short, J. C., Broberg, C., Cogliser, C. C., & Brigham, K. H. (2010). Construct validation using computer-aided text analysis (CATA). An illustration using entrepreneurial orientation. Organizational Research Methods, 13, 320–347.

Simon, H. A. (1962). The architecture of complexity. Proceedings of the American Philosophical Society, 106, 467–482. Simon, H. A. (2002). Near decomposability and the speed of evolution. Industrial & Corporate Change, 11, 587–599. Singh, K. (1997). The impact of technological complexity and interfirm cooperation on business survival. Academy of Management Journal, 40, 339–367. Sirén, C. A., Kohtamäki, M., & Kuckertz, A. (2012). Exploration and exploitation strategies, profit performance, and the mediating role of strategic learning: Escaping the exploitation trap. Strategic Entrepreneurship Journal, 6, 18–41. Sorenson, O., Rivkin, J. W., & Fleming, L. (2006). Complexity, networks and knowledge flow. Research Policy, 35, 994–1017. Stevenson, R. L. (2001). In praise of dumb clerks: Computer-assisted content analysis. In M.D. West (Ed.), Theory, method and practice in computer content analysis (pp. 3–12). Portsmouth, NH: Greenwood. Stuart, T. E., & Podolny, J. M. (1996). Local search and the evolution of technological capabilities. Strategic Management Journal, 17, 21–38. Subramaniam, M., & Watson, S. (2006). How interdependence affects subsidiary performance. Journal of Business Research, 59, 916–924. Teece, D., Rumelt, R., Dosi, G., & Winter, S. (1994). Understanding corporate coherence. Theory and evidence. Journal of Economic Behavior and Organization, 23, 1–30. Tushman, M. L., & O'Reilly, C. A. (1996). Ambidextrous organizations: Managing evolutionary and revolutionary change. California Management Review, 38, 8–30. Uotila, J., Maula, M., Keil, T., & Zahra, A. A. (2009). Exploration, exploitation, and financial performance: Analysis of S&P 500 corporations. Strategic Management Journal, 30, 221–231. Vagnani, G. (2012). Exploration and long-run organizational performance: The moderating role of technological interdependence. Journal of Management, (forthcoming), published online 6 December 2012. Watts, D. J., & Strogatz, S. H. (1998). Collective dynamics of small-world networks. Nature, 393, 440–442. Yayavaram, S., & Ahuja, G. (2008). Decomposability in knowledge structures and its impact on the usefulness of inventions and knowledge-base malleability. Administrative Science Quarterly, 53, 333–362. Zahra, S. A. (1996). Technology strategy and financial performance: Examining the moderating role of the firm's competitive environment. Journal of Business Venturing, 11, 189–219. Zhou, Y. M. (2013). Designing for complexity: Using divisions and hierarchy to manage complex tasks. Organization Science, 24, 339–355.

Please cite this article as: Gatti, C., et al., Interdependence among productive activities: Implications for exploration and exploitation, Journal of Business Research (2014), http://dx.doi.org/10.1016/j.jbusres.2014.07.011