1
Innovation Search: The Determinants of Local and
Distant Search in a Backward-, and Forward-looking
model
MCs Business Administration – Strategy
Name: Anne van ZwietenStudent ID: 11947640
Thesis supervisor: Bernardo Silveira Correia Lima
2
Statement of originality:
This document is written by student Anne van Zwieten who declares to take full responsibility for the contents of this document. I declare that the text and the work
presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and
Business is responsible solely for the supervision of completion of the work, not for the contents.
3
Table of content:
Thesis supervisor: Bernardo Silveira Correia Lima ... 1
Statement of originality: ... 2
Table of content: ... 3
Abstract: ... 4
Introduction: ... 5
Theoretical model and hypotheses: ... 18
Methodology: ... 26
Statistical model: ... 32
Regression analysis: ... 35
Discussion: ... 39
4
Abstract:
This study examines how performance how performance in a backward, and forward-looking model affects decision making about where to search for innovations, i.e. local or distant. Based on a longitudinal analysis of 49 U.S. Chemical firms from 1980 to 1999, this study shows how past performance below aspiration level, and expected performance below target level, determine firms’ search behaviour. This study has developed, and operationalized a forward-looking decision model that translates cognition into organizational behaviour, incorporating prospect theory (Kahneman & Tversky, 1979), and the logic of the Behavioural Theory of The Firm (Cyert & J. March, 1963). Taking into account the bounded rationality constraints of human decision makers, and the rule-based, and goal-directed nature of firms, this model proposes that the comparison between performance targets, and performance expectations guides decision makers and subsequent organizational behaviour in new learning environments. The findings show that both past performance, and expected performance are important determinants of organizational search behaviour. Whereas prior research has suggested that organizations are likely to search in the neighbourhood of their existing knowledge domain when performance falls below aspiration level, findings in the current study show opposing results in that performance below aspiration level increases distant search. In addition, the results reveal that local search is reduced when firms feel unlikely to achieve next year’s performance targets. Instead, distant search is increased when managers expect to underperform the coming year, especially when managers expect to fail relative to the social performance target.
5
Introduction:
Innovation is crucial for economic growth, and organizational adaption (Schumpeter, Opie, & Elliott, 1934). Central to classic and contemporary theories on innovation, is the notion of innovation search (Laursen, 2012; Nelson & Winter, 1982). In innovative search, firms search for new knowledge elements with the ultimate goal of creating new products (Ahuja & Katila, 2004; R. Katila & Ahuja, 2002). Search can be local, which involves search in the vicinity of existing knowledge (Helfat, 1994a; Nelson & Winter, 1982), or distant, which is defined as “a conscious effort to move away from current organizational routines
and knowledge bases” (Katila & Ahuja, 2002, p. 1184). Local search is more reliable and
therefore less risky and cheaper, but the downside of local search is that it frequently lacks the variety, inspiration, and the opportunities needed for new knowledge combinations (Fleming & Sorenson, 2004; Laursen, 2012; Rosenkopf & Nerkar, 2001). Accordingly, local search leads to incremental innovations, i.e. improvement of, or small changes to current
technologies. On the other hand, distant search is carrier more uncertainty and thus involves high costs and more risk. Distant search may lead to the identification of technological disruptions, or radical innovations (Fleming & Sorenson, 2004; Laursen, 2012). If firm succeed to identify knowledge elements that leads to a radical innovations, this in turn can increase performance (Riitta Katila et al., 1998), enable firms to gain competitive advantage (Lopez-Vega, Tell, & Vanhaverbeke, 2016), and empower firms to become the market leader (Mitchell, 1989).
There has been extensive research on how local, and distant search affect the level of innovativeness, performance, and the creation of new products (R. Katila & Ahuja, 2002; Riitta Katila, 2002; Riitta Katila & Chen, 2008). However, a relatively under-researched topic is what triggers firms to engage in local and distant search. Whereas in most decision models manager determine decisions on performance feedback, past experience, and organizational
6 routines (Argote & Greve, 2007; Cyert & J. March, 1963; Greve, 2003; Helfat, 1994a; Levitt & March, 1988; Nelson & Winter, 1982), others have argued that decisions should not only be made as a response to short-term problems, but choice and action should reflect
anticipations of future environments as well (Gavetti & Rivkin, 2007; Greve & Taylor, 2000; Rhisiart, Miller, & Brooks, 2015; Tripsas & Gavetti, 2000). Gavetti and Levinthal (2000) have introduced an organizational behaviour model that comprises a backward-, and forward-looking organizational decision model.
The backward-looking model is drawn on the Behavioural Theory of The Firm (Cyert & J. March, 1963). Organizational behaviour is characterized as history dependent, goal-directed and determined by simple rules (Levitt & March, 1988). Firms set aspirations that serve as a reference point to determine whether performance is satisfactory or not (Cyert & J. March, 1963; Greve, 2003). Aspirations can be a set as a function of prior performance, i.e. the historical aspiration level, or as a function of performance of competitors within the same industry, i.e. the social aspiration level (Cyert & J. March, 1963; Greve, 2003). The
comparison between past performance and aspirations level provide firms with performance feedback (Cyert & J. March, 1963). Performance feedback gives manager a basic idea how well the firm is doing, and whether to search for alternatives to make the current routines change. When firms fail to attain to their aspiration(s), managers are motivated to engage in
problemistic search (Cyert & J. March, 1963; Greve, 2003). The problemistic search
argument suggests that firms increase risk-taking in their search for alternatives to overcome performance shortfalls immediately. Search is myopic, i.e. short-sighted, directing search activities to areas in the vicinity of the pre-existing knowledge (Helfat, 1994a; Nelson & Winter, 1982). In the backward-looking model, firms rely on short-term feedback and neglect the future, because organizations are vulnerable to myopia (Levitt & March, 1988), and expectations are regarded as biased and manipulated within in firm (Cyert & J. March, 1963).
7 The anticipation of the future, or the evaluation of consequences of distant courses of action, is largely absent (Cyert & J. March, 1963; Gavetti & Levinthal, 2000; Gavetti & Rivkin, 2007).
The forward-looking model, on the other hand, reflects decision making based on cognitive representations of the future (Gavetti & Levinthal, 2000). Cognition is defined as a “forward-looking form of intelligence that is premised on an actor's beliefs about the linkage
between the choice of actions and the subsequent impact of those actions on outcomes.”
(Gavetti & Levinthal, 2000, p. 113). Cognitive representations of the future assist managers in finding promising directions of search, which act as templates that guide the subsequent local search efforts (Gavetti & Levinthal, 2000). Prior research on managerial cognition has
revealed the pivotal role of managers’ conceptualization of the organization’s future
environment, and its connection to decision making in a new learning environment. Tripsas and Gavetti's (2000) account on Polaroid’s failure to transition from analog to digital imaging technologies is the perfect example. While Polaroid was able to develop cutting-edge
technologies in digital imaging, senior management decided to move search efforts away from this emerging business, because digitial imagining was inconsistent with Polaroid’s current beliefs, and instead reinforced the current strategy. Accrdingly, Polaroid’s analog business strategy failed to be competitive in the digital imagining world. The Polaroid story has clearly illustrated that the process of organizational search in a new learning environment is deeply connected to the way how managers conceptualize the new problem space, and how decisions of choice and action are made based on this view of the world.
But how are cognitive representaions of the future translated into actual firm
behaviour? Chen (2008) has introduced a behavioural forward-looking model that is drawn on the prospect theory (Kahneman & Tversky, 1979), and in line with the logic of the
8 performance targets, which are the equivalent of aspiration levels in the backward-looking model. The comparison between expected performance relative to future performance targets, provides firms with a basic idea about their future state and determines organizations
subsequent organizational behaviour. Chen (2008) has shown that when firms feel unlikely to achieve target levels, managers’ propensity to risk-taking is likely to increase. In other words, managers increase risk-taking behaviour when they expect to underperform the coming year. Although research and development (R&D) intensity is often used as a proxy for search behaviour, it does not distinguish between local and distant search. Therefore, this study aims to fill this gap and to develop a model that consists of a backward-looking, and a forward-looking decision model in relation to where firms search, i.e. local or distant. Local, and distant search are conceptualized as two orthogonal variables, allowing firms to engage in both types of search simultaneously (Gupta, Smith, & Shalley, 2006; Riitta Katila, 2002)
This study contributes to the current literature in three respects. First, it offers a longitudinal study analysing the data of 49 chemical firms from the period 1980 to 1999 to examine the determinants of local, and distant search in a backward-, and forward looking model. Understanding a firm’s backward, and forward-looking search determinants enables firms to guide search efforts towards innovations, and change. Second, it is not well
understood why firms engage in distant search. This research offers a new perspective, a forward-looking approach, on managers’ motivation to engage in distant search. Third, this study extends the Behavioural Theory of The Firm (Cyert & March, 1963) by adding a operationalized and behavioural forward-looking decision model that translates cognitive search into actual firm behaviour.
This thesis is structured as follows: the first section contains a review of the existing literature on where, and how firms search. This is followed by a theoretical framework including the hypotheses of this study. The methodology section explains the sample, the
9 variables and the statistical tests used in this research. The results section elaborates on the outcomes from the statistical tests. Last, the discussion will discuss the major findings of this thesis, the limitations, and ideas for future research.
10
Literature review:
Where to search
Innovation search can be regarded as a problem-solving activity (Nelson & Winter, 1982), because firms solve problems through combining new knowledge elements with the objective to create new products. In its essence, innovations search can thus be characterized as an organizational learning process (Huber, 1991): search enables firms to improve their current routines (Nelson & Winter, 1982), to develop new skills and learn (Makadok & Walker, 1996), and to adapt to changes in the environment (Cyert & J. March, 1963).
In innovation search, variety is a central theme (Laursen, 2012). Variety is crucial for new knowledge combinations, and to cope with the environmental changes (He & Wong, 2004; Laursen, 2012; Tripsas & Gavetti, 2000). In innovation search, firms can vary their search efforts on two distinct dimensions, i.e. local and distant search. Local search involves search in the neighbourhood of firms’ current knowledge (Fleming & Sorenson, 2004; Helfat, 1994b; Laursen, 2012), and therefore it is cheaper and more cost effective (Laursen, 2012). The downside of local search is that it often lacks the variety, inspiration and the opportunities needed for new knowledge combinations (Fleming & Sorenson, 2004; Rosenkopf & Nerkar, 2001). Accordingly, local search leads to incremental innovations, i.e. advances the existing technology. Local search result in positive term performance effects, though these short-term improvements might come at the expense of long-short-term performance. On the other hand, distant search can be defined as “a conscious effort to move away from current organizational
routines and knowledge bases” (Katila, 2002, p. 1184). Greve (2003) has argued that
successful search efforts are those that are further away from what a firm already knows. Distant search is concentrated on creating variety, and new knowledge combinations (Fleming & Sorenson, 2004; Laursen, 2012; Rosenkopf & Nerkar, 2001), though it carrier more
11 new knowledge combinations (Fleming & Sorenson, 2004) that could possibly lead to the development of radical innovations (Dewar & Dutton, 1986). Radical innovations are regarded as new, and as innovations that “destroy” the current business models (Laursen, 2012). Radical innovations can provide firms with the opportunity to achieve competitive advantage (Y. Chen, Vanhaverbeke, & Du, 2016; Lopez-Vega et al., 2016), and to become market leader (Lopez-Vega et al., 2016).
Now that the linkage between where to search and its pay-offs is clear, the question remains how companies search. The following section will discuss the determinants of local, and distant search as the question of how to search.
How to search
The Behavioural Theory of The Firm (Cyert & J. March, 1963) can be regarded as the starting point of decision-making and organizational behaviour theory. Central to the
behavioural theory is the notion of bounded-rationality (Simon, 1956), which assumes that every outcome of every alternative cannot be precisely forecasted. This belief has given rise to two major research traditions: the evolutionary economics (Nelson & Winter, 1982), and the organizational learning theory (Levitt & March, 1988).
Cyert & March (1963) have suggested that organizational behaviour consists for a big part of procedure following, which imply ‘rules’ for handling information and performing tasks. Rules offer standardization, and simplification of processes, though they are flexible enough to handle variation. Based on this logic, Nelson and Winter (1982) have introduced their theory of organizational routines, which has become a big component of the
evolutionary economics theory. The evolutionary economists have argued that firms are routine-based agents that change incrementally through search (Nelson & Winter, 1982). The
12 evolutionary theory views routines as the basis of stability and organizational continuity, “routines as genes” (Nelson & Winter, 1982, p. 134), is the catch phrase. Routines reflect the experiential wisdom that firms have accumulated over time through trial-and-error learning and the selection and retention of prior behaviours (Levitt & March, 1988). The second trend of research in the behavioural theory is that of the organizational learning theory (Argote & Greve, 2007; Levitt & March, 1988). This theory examines how organizations learn from experience (Levitt & March, 1988), specifically how managers draw inferences from performance (Cyert & J. March, 1963). Cyert and March (1963) have analysed a particular dimension of performance, such as whether it was below or above aspiration levels.
Aspiration levels can be defined as “the minimal outcome deemed satisfactory by the decision
maker” (Schneider, 1992, p. 1053). Managers set aspirations that serve as a reference point to
judge wheter performance is satisfactory or not (Cyert & J. March, 1963; Greve, 2003). Firms set aspiration levels based on prior performance, i.e. historical aspiration level, and as a function of performance of peers within the same industry, i.e. social aspiration level (Cyert & J. March, 1963). The comparison of past performance with aspiration levels, provides firms with performance feedback, and guides subsequent processes of decision making.
Performance feedback gives managers a basic idea how well the firm is doing and whether the organization needs to make the current routines change, and search for alternatives (Cyert & J. March, 1963). Failure to attain performance aspirations motivates managers to engage in
problemistic search (Cyert & J. March, 1963). Problemistic search is defined as “search that is stimulated by a problem (usually a rather specific one) and directed toward finding a solution to that problem” (Cyert & J. March, 1963, p. 121). Failure to attain to the aspiration
levels increases managers’ propensity to risk-taking in their search for alternatives to overcome performance shortfalls immediately (W.-R. Chen & Miller, 2007; Riitta Katila & Chen, 2008). Cyert and March (1963) have argued that search is almost always highly
13 localized, because firms search for solutions in the neighbourhood of the problem. In addition, the evolutinary economists have argued that firms search along established trajectories created by past experience and routines that direct search for alternatives to the vicinity of
organizations’ pre-existing knowledge base (Helfat, 1994b; Nelson & Winter, 1982). Others have proposed that because human decision makers are subject to self-attribution, and
interpretation, they are likely to favour recently performed actions in decision making (James Gardner March, 1991) and therefore the repetition of previous actions is more likely than searching for novelty. Hence, these mechanism give greater short-term rewards to repeated action, than to new ones (D. Levinthal & March, 1981) making local search more lieky than distant search. Furthermore, Cyert and March (1963), have postulated that whereas low performance initially triggers local search, persistent problems lead to expansion of search beyond the confines of the firm. However, Greve (2007) has suggested that the basic
argument of problemistics search, i.e. ‘remedial actions to overcome performance shortfalls’, applies to both exploitation (generally simultaneously used with local search), and to
exploration (equivalent to distant search), and at the same time. Managers that are looking for solutions engage in local search, but also try distant search in their search for a solution. What these theories have in common, is that they are backward-looking. Organizational behaviour is primarily focused on the adaptive adjustments from past performance. Nelson and Winter (2002; p. 29) have argued that: “Learning guided by clear short-term feedback can be
remarkably powerful, even in addressing complex challenges. But that sort of learning does little to enable sophisticated foresight, logically structured deliberation and/or the
improvisation of novel action patterns—and situations that demand these are rarely handled well”. Organizations rely on short-term feedback, without great reference to the future,
because they are vulnerable to myopia (D. A. Levinthal & March, 1993), and because expectations are regarded as biased, and as manipulated within the firm (Cyert & J. March,
14 1963). The impossibility of perfect expectations, calculative presentations of the future is the result of the bounded rationality constraints of human decision makers (Simon, 1956).
Researchers have argued that current technological advances have enhanced organizations’ computation-, and information-processing capabilities (Gavetti & Rivkin, 2007). Indeed, … have observed a trend towards more rational decision making. In this context, researchers have attempted to develop a forward-looking approach, with recognition of managers’ bounded rationally constraints. For example, the real option logic (McGrath, 1997). Real option logic recognizes the behavioural limits of decision makers while providing forward-looking guidelines (Miller & Arikan, 2004), and promoting flexibility in uncertain future environments (McGrath, 1997).
Unlike this rational decision model, others have argued that while human decision makers are boundedly rational, managers do apply a logic of consequences in organizational behaviour (James G. March & Heath, 1994), and do engage in behaviours that are often proactive, cognitive, and deliberate (Porac, Thomas, & Baden-Fuller, 1989; Weick, 1995). Indeed, routines do evolve through processes of local search, because managers are constraint by bounded rationality, however, the “users”, i.e. decision makers may mindfully alter these routines through the decisions managers make. Thus, routines may reflect both emergent and effortful characteristics (Feldman, 2000). Researchers have argued that the notion of bounded rationality itself does not rule out organizational behaviour that is premised on a logic of consequences (March & Simon, 1958; James G. March & Heath, 1994), and have stressed the importance to integrate this logic of consequences with the logic if routine-based organizational behaviour (Perrow, March, & Olsen, 1976).
In this vein, Gavetti and Levinthal (2000) have proposed a behavioural model that comprises two decision models: a backward-looking model based on logic of experience, and
15 the forward-looking logic of consequences. Gavetti and Levinthal (2000) have argued that while decision makers are boundedly rational and indeed cannot perfectly foresee and predict future events, it does not prevent human decision makers from foreseeing possible
consequences of alternative courses of action. Cognition serves as a “forward-looking form of
intelligence that is premised on an actor's beliefs about the linkage between the choice of actions and the subsequent impact of those actions on outcomes.” (Gavetti & Levinthal, 2000,
p. 113). In the forward-looking logic of cognition, firms make cognitive representations of their future decision environment and consider probable consequences within the frameworks of these representations before making decisions (March & Simon, 1958). Subsequent
decisions are based on cognitive representations of the future (Gavetti & Levinthal, 2000).
Gavetti and Levinthal (2000) have suggested that the backward-, and forward-looking model are intimately related and mutually reinforcing. Together they determine organizations’ course of action: “one part [of decision making] occurs in the world of cognition and
comprises ways of conceptualizing the firm and its environment. The other unfolds in the world of action and consists of mechanisms that shape what a company actually does”
(Gavetti & Rivkin, 2007, p. 420). In other words, cognitive representations of the future assist managers in finding promosing directions of search, which act as templates that guide the subsequent local search efforts, and the related emergence of routines (Gavetti & Levinthal, 2000). Which direction of search, and subsequently which path of action to choose, depends largely on how managers conceptualize their future environment.
Prior research on managerial cognition has revealed the pivotal role of managers’ conception of the future environment, and especially how the focal firms perceives it place in it (Weick, 1995). Tripsas and Gavetti's (2000) account on Polaroid’s failure to transition from analog to digital imaging technologies is the perfect example. Despite the absence of a market for digital imaging, and thanks to the early investments in electronic technologies, Polaroid’s
16 technical managers were able to develop leading-edge capabilities in a wide range of
technological areas related to digital imaging. However, Polaroid’s senior managers regarded digital imaging as inconsistent with its current strategy, and moved search efforts away from the emerging digital technologies. By holding on to the current belief structures, and placing primacy on technical excellence, Polaroid was ultimately unsuccessful in the digital world (Tripsas & Gavetti, 2000). This story has illustrated the importance of problem-framing of cognitive representations of the future in decision-making.
But how are cognitive representaions of the future translated into actual firm
behaviour? Chen (2008) has introduced a behavioural forward-looking model that is drawn on the prospect theory (Kahneman & Tversky, 1979), and in line with the logic of the
Behavioural Theory of The Firm (Cyert & J. March, 1963). In this model, firms set future performance targets, which are the equivalent of aspiration levels in the backward-looking model. The comparison between expected performance relative to future performance targets, provides firms with a basic idea about their future state and determines organizations
subsequent organizational behaviour. Chen (2008) has shown that when firms feel unlikely to achieve target levels, managers’ propensity to risk-taking is likely to increase. In other words, managers increase risk-taking behaviour when they expect to underperform the coming year. Although research and development (R&D) intensity is often used as a proxy for search behaviour, it does not distinguish between local and distant search.
This study aims to fill this gap and to develop a model that consists of a backward-looking, and a forward-looking decision model in relation to where firms search, i.e. local or distant. The following research question is developed:
“What are the determinants of local, and distant search, as a function of past performance,
17 Local, and distant search are conceptualized as two orthogonal variables. Following previous studies, this comprises that firms can search on both dimensions simultaneously (Gupta et al., 2006; Riitta Katila, 2002). The model below illustrates the research question and the theoretical framework, which will be discussed in the following section.
-- --
-- --
Where firm search
Forward-looking determinants Expected performance below target level
Local search Backward-looking
determinants Performance feedback of below aspiration level
18
Theoretical model and hypotheses:
Determinants of local and distant search in the backward-looking model
In the backward-looking decision model, problemistic search is triggered when past performance falls below of aspirations (Cyert & J. March, 1963; Greve, 2003). As mentioned earlier, a cornerstone of the Behavioural Theory of the Firm (Cyert & J. March, 1963) is the notion of bounded rationality (Argote & Greve, 2007; Cyert & J. March, 1963; Helfat, 1994b; D. A. Levinthal & March, 1993; Nelson & Winter, 1982; Simon, 1956). In organizational search, firms do not examine all possible alternatives, because bounded rationality limits human decision makers in uncovering all possible options, and in evaluating their future prospects (Helfat, 1994b). The assumption of bounded rationality has its implications for decision-making and the subsequent direction of search.
First, perhaps the most characteristic of organizational learning is that it is cumulative. Cohen and Levinthal (1990) have suggested that “learning performance is greatest when the
object of learning is related to what is already known” (p. 131). The cumulative nature of
learning has implications for organizational search. In searching new knowledge elements, firms tend to search for new knowledge that is closely related to the old knowledge. Hence, firms tend to focus on areas related to the current knowledge domains, especially on areas close to the organization’s current expertise (Helfat, 1994a; Levitt & March, 1988). Firms tend to concentrate their search efforts in areas related to the current knowledge, because learning is easier, and more cost effective if it is restricted to familiar and proximate neighbourhoods (Cohen & Levinthal, 1990). Firms learn which components have failed in previous inventions and have stopped using them. Thus, firms learn to avoid the combinations and architectures that have failed in the past (Fleming & Sorenson, 2004; Vincenti, 1990). Moreover, firms develop an understanding of their “local” environment and recognize elements that potentially could be combined, and recombined (Fleming & Sorenson, 2004).
19 Thus, firms tend to concentrate their search efforts towards areas that are close to the pre-existing knowledge domain, because this type of search is easier, more reliable, and cost effective (Fleming, 2001).
Second, the evolutionary economists view routines as a feature of continuity, stabilization in firm behaviour, and search is regarded as a source of organizational change (Helfat, 1994b; Nelson & Winter, 1982). If routines and standard operating procedures are relatively stable over time, firms strongly rely on prior choices, thus reinforcing any persistence in R&D as a result of cumulative learning (Argote & Greve, 2007; Cyert & J. March, 1963; Nelson & Winter, 1982). Persistence in R&D reinforces the tendency to pursue activities similar to those undertaken in the past. Thus, persistence in R&D produces patterns of search efforts that reflect continuing activities in the same technological direction as in the past. Hence, when performance falls below the aspiration level, organizational change will become more likely. Because problemistic search is myopic, i.e. short-sighted, the changes are likely to occur near in areas that the firm has recently changed in, hence in areas near to the apparent problem.
Third, Levinthal and March (1993) have suggested that organizations are vulnerable to spatial myopia. That is, learning tend to favour the effects that occur near to the learning, and ignore the larger picture. Firms do not consider all the options, but they are usually satisfied with the first option. Hence, when searching for solutions, firms choose the first alternative available that satisfies aspirations levels and helps firms to overcome performance shortfalls. The argument have led to the following hypothesis:
H1a. Past performance below of aspiration level (social or historical) increases local search efforts.
20 Furthermore, Greve (2007) has suggested that the problemistic search argument, i.e. "remedial actions to overcome performance shortfalls’, applies to both local search, and to distant search, and at the same time. In this context, firms can not only choose to which extent to which they seek to innovate, but also whether to emphasize incremental innovations
through local search, or direct search towards finding radical innovation through engaging in distant search.
Whereas the previous section has argued that organizational routines facilitate learning closely related to the pre-existing knowledge domain, Jelinek and Schoonhoven (1990) have argued that organizations can also develop routines for making exploratory innovations, in other words routines to distant search.
In addition, Rosenkopf and Nerkar (1999) have suggested that firms’ innovation trajectories are affected by technological actions of other firms within the same industry. Additionally, Rosenkopf and Tushman (1998), and Tushman and Rosenkopf (1992) have proposed that the technological evolution of a product is generated by communities or firms within in the same industry. Hence, the social and competitive relation of the organization and its peers has a modifying and catalytic effect on the rate of innovation of the focal firm. Thus, the interdependent evolution of firm trajectories suggests that if a firm observes other firms within the same industry making radical innovations, the focal firm is likely to follow and to increase distant search, in its search for new knowledge elements and hence new
combinations.
Moreover, search activities that are further away from the firm’s current knowledge base, have less known consequences and can therefore be regarded as more risky. Prior research has indicated that performance below aspiration level causes increased risk-taking
21 (Tversky & Kahneman, 1991) making distant search an attractive alternative in managers’ search for solutions. Thus, the above mentioned references lead to the following hypothesis:
H1b. Past performance below of aspiration level (social or historical) increases distant search efforts.
Determinants of local and distant search in the forward-looking model
In the forward-looking model, decision are based on cognitive representations of the future (Gavetti & Levinthal, 2000). But how do firms translate cognition into actual firm behaviour? I have developed and operationalized a forward-looking decision model that is line with rule-based, and goal-directed nature of the Behavioural Theory of the Firm (Cyert & J. March, 1963), and that takes into account the logic of the prospect theory (Kahneman & Tversky, 1979). Kahneman and Tversky (1979) have proposed that decision makers are affected by the framing of the future state compared to the status quo, in terms of likely gains or losses. This framing simplifies decision-making, with recognition of bounded rationality of human decision makers. This study proposes that decisions and organizational behaviour in the forward-looking model is determined by the comparison of expected performance to performance targets. Firms set performance targets (analogous to aspirations in the backward-looking model), which serve as a proxy to evaluate expected performance in terms of success or failure. Performance expectations reflect a forward-looking cognitive imagine, because the future environment is taken into consideration and the choices of actions and their pay-offs. Firms frame their future state as “likely outperforming”, or “likely underperforming”, and decision making and subsequent organizational behaviour. This framing presents a simplified decision rule that recognizes the bounded rationality constraints of human decision makers in predicting a set of possible choices of action and the possible consequences.
22 Gavetti and Levinthal (2000) have suggested that applying a forward-looking
intelligence, ‘dramatically’ enhances firms’ adaptive behaviour. Gavetti and Levinthal (2000) have theorized that cognitive representations of the future help firms to identify promising areas on the ‘fitness landscape’, which are then carved out through a process of experiential search. However, in business practice, at current times of environmental uncertainty, decision makers are faced with high levels of complexity, and discontinuity. Fruitful directions of search and the appropriate paths of actions are therefore rarely obvious in the moment of decision making. For example, Vincent's (1994) account on retractable landing gear has illustrated that this new type of landing gear for airplanes was not directly recognized as a discontinuity by many in the industry and as a potential dominant design as compared to the ‘pants’-type fixed landing gear. Similarly, Rosenbloom and Cusumano (1987) case study on the Betamax–VHS video tape battle has shown that, whereas the majority regarded the video tape was important, the specific benefits, and the true nature of the discontinuity were not apparent until later. These two examples have illustrated that all of these factors were known
ex post, but unknown ex ante. In this vein, Weick (1995) has highlighted that managers need
to ‘make sense’ of their environment, before they can act. Hambrick and Mason (1984) have argued that cognitive representations of the future may play a significant role in shaping decision makers’ responses (Hambrick & Mason, 1984). In situations of uncertainty,
especially pervasive uncertainty, i.e. “situations where neither all the different outcomes, nor
their probabilities are initially known, and which are so ill structured that the possible outcomes will remain unknown despite any increase in information routines” (Becker, 2004,
p. 657), routines make important contribution to the manager’s decision of course of action (Gersick & Hackman, 1990; Scapens, 1994; Weiss & Ilgen, 1985). The link between uncertainty and routines is best described by Heiner (1983), who wrote that “. . . greater
23
regularities, so that uncertainty becomes the basic source of predictable behaviour” (p. 570).
In decision making, managers rely on organizational routines in order to reduce uncertainty and therefore firms are more likely to search for knowledge that is more familiar, and related to the pre-existing knowledge base. When firms expect to underperform in uncertain
environments, managers are likely to reduce uncertainty and rely on the exiting routines.
In addition, Tripsas and Gavetti's (2000) account on Polaroid which I described earlier, has illustrated that those managers that interpret innovations in their future environment as a threat to their current strategy, tend to interfere in the change process and reinforce the current strategy instead (Dutton & Jackson, 1987; Staw, Sandelands, & Dutton, 1981), making the continuation of the current, thus local search activities more likely.
Furthermore, to continue with Polaroid, Gavetti (2005) has highlighted the role of organizational hierarchy in relation to decision-making. Whereas the digital-imaging managers started to question the strategic implications of digital imaging and subsequently embraced a business model that centralized hardware, senior managers rejected this emerging business model because it was inconsistent with their believe in the current strategy. The lack of coordination between departments that were located differently in the organizational hierarchy, has led to the failure of senior management ability to revise the current beliefs, or to change the perceptions, when the outlying environment demanded it and reinforced the current strategy. In contrast, Burgelman's (1994) case on Intel has shown that despite the fact that senior management failed to recognize a number of major shifts in the firm’s
environment, Intel was able to manage a major discontinuity in its business, because the senior management team did not interfere with autonomous decisions generated at the lower levels of the firm with their focus on the local market. These two examples have illustrated the role of organizational hierarchy on decision making, espcially with regards to the
24 of skills and knowledge between different hierarchical levels, make it more likely that
decision makers are ignorant of the exogenous changes and therefore are likely to reinforce the current strategy and continue local search efforts. Based on previous arguments the following hypothesis is proposed:
H2a. Expected performance below of target level (social or historical) increases local search efforts.
Moreover, innovations are both competitive, and cognitive challenges (Greve & Taylor, 2000), suggesting new ways of competition and threat to the established order of competition. Innovations signals to managers that the current organizational routines may be inadequate, which creates problems for the organization (Greve & Taylor, 2000). Firms establish routines over time to handle typical environmental changes (Nelson & Winter, 1982), but lack the patterned responses to unfamiliar environmental events (Barr, 1998). Routine responses are regarded as inadequate, because innovations are novel, causing firms to search for alternatives through non-routine search (Cyert & J. March, 1963), probing for knowledge that is beyond the boundaries of the firm. Greve and Taylor (2000) have suggested that observing innovations of peers, increases managers’ motivation to search, and change. On the one hand, innovations hurt the focal firm’s market shares which causes the focal firm to fall below their status quo (Kahneman & Tversky, 1979), increasing managers’ propensity to risk-taking behaviour and tendency to engage in distant search that lead to making changes (Greve, 1998; Greve & Taylor, 2000). On the other hand, innovations by others are perceived as a threat to the organization's current position, and trigger action. M.-J. Chen and Miller (1994) have suggested that firms generally respond to challenges in their market by making competitive counterattacks, so innovations by others are likely to lead to search for
innovations, thus search efforts that are further away from the firm’s pre-existing knowledge. Moreover, seeing others adopt innovations, increases managers’ motivation to search in the
25 same direction, because the fact that other firms have adopted technologies in certain areas that are successful senses the belief that it will be profitable, based on the information available to decision makers (Rogers, 1995; Strang & Soule, 1998). This increases the estimated value for managers who have little information about the consequences
(Bikhchandani, Hirshleifer, & Welch, 1992). Observing other firms introducing successful innovation that are the result of distant search, increases managers’ confidence that this direction of search will be successful. The above arguments have led to the following hypothesis:
H2b. Expected performance below of target level (social or historical) increases distant search efforts.
26
Methodology:
This research was conducted in the chemicals industry (SIC 28). I believe this industry is a suitable setting for this research because it is a diverse market, with high technological intensity (Klevorick, Levin, Nelson, & Winter, 1995). Market diversity in the chemical industry is the result of local differences in downstream products, the need for technical assistance, customer bases, and environmental regulations (Landau & Arora, 1999; Rosenberg & Aurora, 1998). In addition, the chemical industry is technology intensive (Klevorick et al., 1995). To illustrate this statement with an example: the U.S. chemical industry invested in 1992 $16,7 billion in R&D, which was more than any other industry in the U.S. (Rosenberg & Aurora, 1998). Accordingly, patenting is important in de chemical industry (Levin, Klevorick, Nelson, & Winter, 1987). Firms use patents to protect their intellectual property. This enables us researchers to use this patent information to model the innovative behaviour of these firms. The link between patents and innovation is likely to be stronger in industries in which patents provide firms with fairly strong protection for their proprietary knowledge. Prior research has indicated that the chemicals industry is such a particular industry in which patents are
generally used widely and consistently and regarded to be effective, relative to most other industries (Levin et al., 1987). In general, using patent data is convenient because it comprises chronological, consistent information that provides an explanation how firms search, and solve problems over time (Riitta Katila et al., 1998; Riitta Katila & Ahuja, 2002).
Data
This study uses a longitudinal dataset comprising the patenting activities of 49 leading firms from the chemicals industry in the United States (SIC 28). To ensure the availability and reliability of data, the sample was selected to include the largest chemicals firms, which constitute the core of the global chemicals industry and leading firms from longitudinal
27 patenting information from the period 1980 until 1999. The original sample consisted of 1100 firms. Then, 651 firms were dropped because they had less than ten yearly observations. From the remaining firms, 400 of these firms were dropped, because they less than 10 patents. The final sample consisted of 49 firms, the key players in the industry over the study period.
This study obtained yearly patent counts, and firm-attribute data for the firms in the sample. The panel used for the analysis includes patenting activity for the period 1980 until 1999. I used U.S. patent data for all firms in the sample, to maintain consistency, reliability, and comparability. The patent citations are retracted from the NBER U.S. Citation Patent File (Hall & Jaffe, 2001). This database comprises all utility patens in the USPTO’s database granted from the period 1963 to 1999.
The financial figures are retrieved from two major databases, Wharton CRSP and Compustat. They are combined by year, and a GVKEY number is assigned to each firm.
Variables:
Dependent variables:
Local search: Drawing on Katila and Ahuja's (2002) model of search, “search depth”
is conceptualized as local search. Search depth corresponds to the accumulation of search experience with the same knowledge elements (R. Katila & Ahuja, 2002), which is closely related to the definition of local search, i.e. the search for knowledge that is in the
neighbourhood of the current knowledge base. Following Katila and Ahuja (2002), local search is measured by “the average number of times a firm repeatedly used the citations in the
patents it applied for” (p. 1187). The local search variable is created by calculating the
average number of times that a citation in year t-1 was used repeatedly during the past five years. This is consistent with the following formula:
28
Distant search: Katila and Ahuja (2002) have conceptualized “search scope” as the
exploration of new knowledge (R. Katila & Ahuja, 2002). Therefore, for this study search scope is used to measure distant search, i.e. the search for knowledge that is beyond the boundaries of the firm. Distant search is measured by the proportion of previously unused citations in a firm’s focal year’s list of citations. Distant search is calculated by the number of citations in a focal year’s citations that could not be found in the list of patents and citations by that firm over the previous five years. This is summarized in the following formula:
Independent variables:
Determinants of backward-looking search: In the backward-looking model past
performance is measured by the accounting performance measure return on assets (ROA) (net income divided by total assets). Performance levels are evaluated against firms’ social and historical aspiration level (Cyert & J. March, 1963; Greve, 2003). The comparison between performance and aspiration level serves as a reference point to determine whether
performance is satisfactory or not (Greve, 2003). Performance can be measured as a function of its own prior performance (historical aspiration level), and relative to its peers (social aspiration level). Following Chen and Miller (2007), the variables for the aspiration levels were lagged with one year. The historical aspiration level is calculated at the average ROA of
29 the focal firm over the past three year. The social aspiration level is calculated as the mean ROA of all companies within the same chemical industry. Performance feedback which is the determinant of subsequent search behaviour, is calculated at performance minus aspiration level at a proxy of t-1. In order to test the different effects between performance above, and below aspiration level, spline variables were created (Desai, 2008). Spline variables were created by separating performance below, and above aspiration level (Greve, 2003). Performance of below aspiration level only has observations that are below zero, and the performance of above aspiration those that are greater than zero .
Determinants of forward-looking search: In the forward-looking model, expected
performance is measured by Tobin’s q ratio. Tobin’s q is a market-based measure which captures the long-term prospects (Bharadwaj, Bharadwaj, & Konsynski, 1999; Harrigan et al., 2018). Long-term prospects reflects performance expectations because it takes into
consideration the future environment and the choices of actions and their pay-offs. Tobin’s q is calculated by dividing the market value of total assets, by the book value of the firms’ total assets (Dybvg & Warachka, 2013). The Tobin’s q ratio indicates whether investors expect high growth opportunities or low ones (Harrigan et al., 2018). In line with the goal-directed, and rule-based nature of the Behavioural Theory of The Firm (Cyert & J. March, 1963), performance expectations are measured compared to the focal firm’s target level (historical target), and compared to its peers (social target level). Both target levels were lagged with one year. Whether expected performance is positive, or negative is calculated by expected
performance minus target level (historical/ social) at t-1. This outcome determinants firms’ consequent actions of search behaviour. Equivalent to the backward-looking decision model, in order to test the different effects between expected performance above, and below target levels, spline variables were created (Desai, 2008). Spline variables were created by
30 separating expected performance below, and above target level (Greve, 2003). The
expectancy to underperform only has observations that are below zero, and the expectancy to outperform only those that are greater than zero.
Control variables:
Firm size: The size of a firm may affect firms in decision making in terms of risk
taking, and innovation search decisions (Lim & McCann, 2014) and for this reason I control for firm size. The variable is constructed as the number of logged employees in a firm, which is an appropriate measure for the overall firm size within the industry (Audia & Greve, 2006). This control variable is calculated by the logging the number of employees in a given firm. Following Audia and Greve (2006a), a logged number of employees ensures that an increase in employees, does not affect firm decisions.
Slack: As slack resources may influence firm decisions in terms of where to search, I
have controlled for slack (Argote & Greve, 2007; Greve, 2003). Two measures of slack that were used in previous studies I have controlled for: unabsorbed slack, and potential.
Following Chen and Miller (2007), unabsorbed slack is calculated using working current ratio (current assets-to-current liabilities ratio). In addition, potential slack was measured as the ratio to dept to equity (Wiseman & Bromiley, 1996).
Distance to bankruptcy: Since March and Shapira (1992) have shown that firms are
more averse to risk when facing bankruptcy, which may influence firms’ search decisions, I have controlled for risk aversion by including the Altman’s Z measure. Following previous studies (e.g. W.-R. Chen & Miller, 2007; Miller & Chen, 2004) the distance from bankruptcy using Altman’s Z-score is measured by: “(1.2 x working capital divided by total assets) + (1.4
31 x retained earnings divided by total assets) + (3.3 x income before interest expense and taxes divided by total assets) + (0.6 x market value of equity divided by total liability) + (1.0 x sales divided by total assets).” A higher Altman’s Z-score value suggests a lower risk of
32
Statistical model:
I ran the Breusch-Pagan Lagrange multiplier test to check for serial correlation. The test very soundly rejected the null hypothesis for both local search: (𝜒𝜒2(8) = 358.94, p <
0.01), and distant search: (𝜒𝜒2(8) = 124.78, p < 0.01). This indicates that there are individual
firm effects and therefore the simple OLS estimator is not an adequate model for this study. In order to determine whether to use the within (fixed effects), or GLS (random effect) estimator (Mutl & Pfaffermayr, 2011), I ran the Hausmann specification test. The result indicated significant results for both local search (𝜒𝜒2(8) = 19.00, p < 0.001), and distant search (𝜒𝜒2(8) =
87.28, p < 0.001). Therefore, the fixed effects method is chosen to model the data instead of the random effects model.
Table 1 provides an overview of the descriptive statistics and correlations for the independent, and control variables for local search, and distant search. The sample consists of 910 observations from 1980 until 1999.
First, the following paragraph considers the correlations of the control variables, and the independent variables for local search. Table 1 presents that past performance below the social aspiration level (r = 0.04, p > 0.05), and below the historical aspiration level (r = 0.02, p > 0.05) both have a positive, but insignificant correlation with local search. On the other hand, expected performance below the social target has a positive, and not significant correlation with local search as well (r = 0.06, p > 0.05). In contrast, expected performance below the historical target level is negatively correlated with local search (r = -0.08, p < 0.05)
Moreover, with regards to the control variables, firm size is negatively correlated with local search (r = -0.14, p < 0.05). Potential slack (r = 0.1, p > 0.05), unabsorbed slack (r = 0.18, p > 0.05), and distance to bankruptcy (r = 0.14, p < 0.05) show a small, and positive correlations with local search.
33 Last, the past performance below historical aspiration level, and past performance below social aspiration level variables are highly correlated with each other (r = 0.89, p < 0.05). Therefore following prior studies, I ran separated models for the past performance below historical, and social aspiration level to avoid distorted parameter estimates (e.g., Chen & Miller, 2007; Iyer & Miller, 2008; Lim & McCann, 2014).
Second, this paragraph explains the correlations of the control, and independent variables, with distant search. It shows that past performance below the social aspiration level, has a positive, but insignificant correlation with distant search (r = 0.05, p > 0.05). Additionally, past performance below the historical aspiration is positively correlated with distant search (r = 0.07, p < 0.05). On the other hand, expected performance below target the social target level has a small, and negative correlation with distant search (r = -0.14, p < 0.05). Contrarily, expected performance below the historical target level is positively
correlated with local search (r = 0.16, p < 0.05). Last, considering the control variables, firm size is positively correlated with distant search (r = 0.31, p < 0.05), whereas potential slack (r = 0.09, p < 0.05), unabsorbed slack (r = 0.21, p < 0.05), and distance to bankruptcy (r = -0.18, p < 0.05) are negatively correlated with distant search.
34 Table 1: Descriptive statistics determinants of local and distant search
N Mean S.D. 1 2 3 4 5 6 7 8 9 10 11 12 13 1. Local search 910 0.64 0.55 2. Distant search 910 0.55 0.19 -0.6* 3. Firm size 910 8.88 2.1 -0,14* 0.31* 4. Depth to equity 910 0.55 6.14 0.1* -0.09* -0.08* 5. Unabsorbed slack 910 2.87 6.14 0.18* -0.21* -0.55 -0.04* 6. Distance to bankruptcy 910 2.88 3.77 0.14* -0.18* -0.38* -0.09* 0.58*
7. Past performance below
aspiration level (social) 910 -0.3 0.12 0.04* 0.05 0.24* 0.01 -0.07* -0.01 8. Past performance above
aspiration level (social) 910 0.2 0.24 0.11* -0.06 0.13* -0.03 -0.04 0.1* 0.17* 9. Past performance below
aspiration level (historical) 910 -0.04 0.12 0.02 0.07* 0.27* -0.01 -0.12* -0.02 0.89* 0.16* 10. Past performance above
aspiration level (historical) 910 0.03 0.07 0.12* -0.14* -0.19* 0.08* 0.06 0.16* 0.02* 0.11* 0.14* 11. Expected performance
below target level (social) 910 -0.97 1.14 0.06 -0.14* 0.01 0.05 -0.05 0.06 -0.06 -0.24* -0.01 0.02 12. Expected performance
above target level (social) 910 0.31 1.13 0.13* -0.13* -0.14* 0.1* 0.04 0.33* -0.03 0.14* 0.01 0.21* 0.23* 13. Expected performance
below target level (historical) 910 -0.26 0.75 -0.08* 0.16* 0.3* 0.03 -0.22* -0.2* 0.04 -0.07* 0.09* -0.01 0.02 -0.06 14. Expected performance
above target level (historical) 910 0.4 1.01 0.14* -0.14* -0.12* 0.17* 0.09* 0.33* -0.06 0.17* -0.05 0.19* 0.13* 0.73* 0.16*
35
Regression analysis:
Table 2 and 3 present the results of the within effects regression analyses of local, and distant search. Table 3 reports the results of local search, and table 4 shows the outcomes for
distant search. Each table consists of a baseline model, this one for local, and one for distant
search, followed by two models for past performance (social/ historical), and two for expected performance (social/ historical).
The results are interpreted following Greve (2003, 2007). Greve (2003, 2007) has shown that when performance drops below of aspirations, a negative coefficient implies an increase in search, whereas a positive coefficient suggests a decrease in search.
First, the results of the regression analysis of past performance on local, and distant search will be analysed. Hypothesis 1a predicted that past performance below of aspiration level (historical/ social), enhances local search. In table 3, model 2, the coefficients for past performance below social aspiration level (𝛽𝛽 = 0.01, p > 0.05), and table 3, model 2, the
coefficient of past performance of below historical aspiration level (𝛽𝛽 = 0.03, p > 0.05) are
both positive, though insignificant. These findings suggest that hypothesis 1a did not receive support.
Hypothesis 1b proposed that past performance below of aspiration level (historical/ social), increases distant search. Table 3, model 7, presents that the coefficient for past
performance below social aspiration level, is negative and significant (𝛽𝛽 = -0.15, p < 0.05). In
addition, the coefficient for performance below of the historical aspiration level, table 3, model 8, is significant and negative as well (𝛽𝛽 = -0.11, p < 0.05). These results suggests that
firms increase distant search when they fail to attain to their aspiration level(s). Thus, hypothesis 1b is supported.
36 Second, the outcomes of expected performance relative to firms’ target level, on local and distant search will be examined. Hypothesis 2a postulated that expected performance below of target level (historical/ social), increases local search. The coefficients for expected performance below historical target, in table 2, model 5 (𝛽𝛽 = 0.07, p < 0.05), and below social
target, table 3, model 3 (𝛽𝛽 = 0.03, p < 0.05) are consistent, they are both positive, and
significant. This suggests that firms decrease local search, when they expect performance to drop below of their social, and historical target level. These findings suggest that hypothesis 2a rejected.
Hypothesis 2b estimated that expected performance below of target level (social/ historical), enhances distant search. Table 3, model 10, presents the coefficient of expected performance below of historical target level, which is positive, though insignificant (𝛽𝛽 = 0.01,
p > 0.05). Contrarily, table 3, model 9, represents a negative, significant result (𝛽𝛽 = -0.02, p <
0.05) for expected performance below the social target level. This suggests that firms increase distant search, when they expect to underperform relative to their social aspiration level.
37 Table 2: Results of the fixed effects regression for
local search
Model 1 Model 2 Model 3 Model 4 Model 5
Variables β SE β β SE β β SE β β SE β β SE β
Firm Size 0.2** 0.04 0.2** 0.04 0.2** 0.04 0.17** 0.04 0.16** 0.04
Potential slack 0.01** 0.002 0.01** 0.002 0.01** 0.002 0.01* 0.002 0.01** 0.002
Unabsorbed Slack -0.01 0.01 -0.001 0.01 0.002 0.01 -0.002 0.01 0.004 0.01
Distance to Bankruptcy 0.01 0.003 0.01 0.003 0.01 0.004 0.004 0.003 0.01 0.004
Past performance below Social Aspiration 0.01 0.19
Past performance above Social Aspiration 0.29** 0.07
Past performance below Historical Aspiration 0.03 0.18
Past performance above Historical Aspiration 0.45 0.28
Expected performance below Social Target 0.03* 0.02
Expected performance above Social Target 0.05 0.02
Expected performance below Historical Target 0.07* 0.03
Expected performance above Historical Target 0.06** 0.02
F 8.89 8.17 5.45 7.2 7.45
R² 0.01 0.003 0.01 0.01 0.01
38 Table 3: Results of the fixed effects regression for
distant search
Model 6 Model 7 Model 8 Model 9 Model 10
Variables β SE β β SE β β SE β β SE β β SE β
Firm Size -0.02† 0.01 -0.02 0.01 -0.02 0.01 -0.02 0.01 -0.02 0.01
Potential slack -0.001 0.001 -0.001 0.001 -0.001 0.001 -0.0003 0.001 -0.001 0.001
Unabsorbed Slack 0.02** 0.001 0.01** 0.003 0.01** 0.003 0.01** 0.003 0.01** 0.003
Distance to Bankruptcy -0.003** 0.001 -0.002 0.001 -0.002† 0.001 -0.001 0.001 -0.001 0.001
Past performance below Social Aspiration -0.15** 0.06
Past performance above Social Aspiration -0.1** 0.02
Past performance below Historical Aspiration -0.11* 0.05
Past performance above Historical Aspiration -0.13 0.09
Expected performance below Social Target -0.02** 0.01
Expected performance above Social Target -0.02** 0.01
Expected performance below Historical Target 0.01 0.01
Expected performance above Historical Target -0.03** 0.01
F
9.02 9.56 6.22 9.27 7.15
R² 0.06 0.04 0.06 0.03 0.5
39
Discussion:
This study examines how past performance, and performance expectations affect decision making about where to search for innovations, i.e. local or distant. Based on a longitudinal analysis of U.S. Chemical firms from 1980 to 1999, this study shows how past performance below aspiration level, and expected performance below target level, determine firms’ search behaviour.
The findings have shown that both past performance, and expected performance are important determinants of organizational search behaviour. Whereas prior research has proposed that firms almost always search in the vicinity of the problem space, the results of this study have revealed that performance below aspiration level increases managers’ motivation to engage in distant search. This finding reflects Greve’s (2007) argument that problemistic search applies not only to local search, but to distant search as well. In line with Greve (2007), managers solving problems do turn to local search as a solution, but also try distant search. This is different from current theory, which specifies that low performance initially triggers local search, and persistent problems cause expansion of search beyond the confines of the firm (Cyert & J. March, 1963; Riitta Katila & Ahuja, 2002).
Second, choice and action in organizations are not only reactive, and rule-based. They are not purely action-response processes that react to short-term problems, or failures to achieve aspiration levels that have been set in the past. Important decisions often result from deliberate attempts to anticipate future environments. Gavetti and Levinthal (2000) have conceptualized a forward-looking model of cognitive search. Decision makers form cognitive representations of the future and decisions are determined based on the believe in certain links between choice of action, and the outcome of these actions. However, Gavetti and Levinthal (2000) do not specify how firms translate cognitive search into actual firm behaviour. In this study, I have developed and operationalized a forward-looking decision model in line with the
40 rule-based, and goal-directed nature of The Behavioural Theory of The Firm (Cyert & J. March, 1963) that incorporates the logic of the prospect theory (Kahneman & Tversky, 1979), and with recognition of bounded rationality of human decision makers (Simon, 1956). I have proposed that firms set future targets, that serve as a reference point to evaluate performance. Performance expectations reflect a simplified cognitive representation of the future that takes into consideration the conceptualization of a future environment, choices of actions, and their possible outcomes. The comparison between performance target, and performance
expectations determine firms’ subsequent behaviour. Firm frame their future state as “likely underperforming”, or “likely outperforming”. This framing is a simplified decision-rule that guides managers in their decision making regarding where to search, i.e. local or distant.
The analyses of this research have shown that when firms expect to fail to achieve their performance target the coming year, local search is likely to decrease. Instead, my findings have revealed that decision makers’ motivation to engage in distant search is increased when managers expect to underperform relative to their social target the coming year. This result has suggested that competition increases managers’ motivation to move their search efforts away from their current beliefs and strategy, and instead direct their search efforts beyond the confines of the firm. This is in line with Greve and Taylor's (2000) finding that observing innovation of others increases search, and change. Innovations of peers hurt the focal firm’s market shares, which causes the focal firm to fall below their status quo
(Kahneman & Tversky, 1979), increasing managers’ propensity to risk-taking behaviour (W.-R. Chen, 2008) and tendency to engage in distant search (Greve, 1998; Greve & Taylor, 2000). Moreover, innovation by others is perceived as a threat to the organization’s current position and in line with Miller and Chen (2004) firms are likely to respond to threats, or challenges in their market, by making competitive counterattacks by increasing search in distant areas in their search for new knowledge combination. From another perspective, in
41 line with Rogers (1995) and Strang and Soule (1998), observing other firms having adopted certain technologies that are successful in certain areas, senses the belief that this a certain direction of search will be profitable. This increases the estimated value for managers who have little information about the consequences of distant search efforts (Rogers, 1995; Strang & Soule, 1998).
Overall, decision-making has given way to learning, routines and increased focus on change and adaption. This study has highlighted how framing of history and future determine behaviour. Managers can direct organizational change or innovation, by choosing a correct level of aspiration or target. Rather than just responding to performance feedback, managers can take proactive action by making projections about the future and preparing for it.
This study has several limitations that suggest a need for further work. First, Patel and Pavitt (1997) have suggested that patens measure codified knowledge, but don’t measure tacit knowledge. A high proportion of firm-specific competencies, such as R&D expenditures related to search activities, imply non-codified (i.e. tacit) knowledge. In turn, whereas distant search implies the search for new innovations, which will be recorded in terms of patents, local search implies refinement search and often does not lead to patents, which is not codified in patents, which patents fail to measure. Therefore I cannot be sure that the
insignificant of hypothesis 1a: the effect of performance below aspiration on local search, and hypothesis 2a: positive effect of expected performance below target level on local search, are because the effects are not there, or because the effects are too weak to be distinguished in the data. It suggests the need for future research using complementary approaches to really understand how firms search, in relation to where firms search. For example, through case studies. Second, the single-industry research design leaves doubts about the generalizability of the findings. Previous studies have shown that the propensity for patenting varies
42 investigation I cannot be sure that the organizational search processes in my studies applies to other industries as well.
Future research could further examine the backward-, and forward-looking
determinants of local and distant search by incorporating slack resources. Previous research has shown that slack influences R&D investment decisions. Future studies could research how slack affects the determinants of local, or distant search in a backward-, and forward-looking model. Researchers could also investigate if there are differences in the determinants of local, and distant search between big, and small firms. Moreover, another suggestion for future research would be how the determinants of local, and distant search in a backward-, and forward-looking approach of incumbent firms, are affected by start-up firms. Last, but not least, a backward-, and forward looking model of search could also be applied studies
traditionally dominated by the backward-looking rationale, such as studies on organisational decision making, strategic reorientation, or change.
