Learning or inertia? The impact of experience and knowledge codification on post-acquisition integration

Learning or inertia? The impact of experience and knowledge codification on post-acquisition... Abstract This article develops and tests a theory on the evolution of complex organizational decisions, such as the decision to integrate (or not) a target company during the post-acquisition management phase. Using a sample of US bank mergers, we show that persistence in—or variation of—integration decisions depends on two key factors: integration experience and related knowledge codification. Integration experience tends to generate persistence in the integration decision, which is associated with poor deal performance. Moreover, we show that when knowledge codification is low, the persistence caused by integration experience further increases. However, when knowledge codification is sufficiently high, the inertial effect of experience diminishes significantly. Hence, high levels of knowledge codification can weaken the effects of decision inertia, which suggests that as knowledge codification increases, the role of knowledge codification switches from strengthening inertia to promoting learning. 1. Introduction Ample empirical evidence suggests that as an organization gains experience with a particular decision, it becomes more likely to repeat it (Amburgey et al., 1993; Gulati, 1995; Shaver et al., 1997; Dimov et al., 2012). Underlying this perspective is the assumption that experience facilitates the formation and refinement of routines (Levitt and March, 1988). Routines are programs of action that, reflecting an organization’s prior experience with a particular decision or task (Nelson and Winter, 1982), may constrain the organization’s future behavior (March and Simon, 1958; Cyert and March, 1963) and generate inertial decision-making (Szulanski, 1996). This may also happen in the context of strategic decisions, where routines often “play a crucial rule in the formulation of a firm’s strategic choices by supplementing, or even substituting for, calculative, formal strategic decision-making rules” (Haleblian et al., 2006: 358). In this article, we first closely follow the prior literature and establish our baseline hypothesis, arguing that as organizational experience with a given decision increases, the likelihood of repeating that decision in the future increases. When persistence in decision-making is caused by inertia, the stability-proving effect of routines can develop into a pathology (Becker, 2004). Based on the understanding that persistence in decision-making can be harmful, we therefore study how knowledge codification—the codification of related knowledge in written tools such as manuals, blueprints, spreadsheets, and decision support systems (Zollo and Winter, 2002)—may affect this persistence. Note that whereas knowledge codification and experience are likely to covary, they represent two different concepts because “not all codifiable … knowledge is actually codified” (Zollo and Singh, 2004: 1238). Therefore, firms with equal levels of experience might have different levels of knowledge codification (Romme et al., 2010). There are various perspectives on the possible effects of codification. A stream of literature contends that knowledge codification results in true organizational learning, which should remedy the inertial persistence in decision-making. In particular, knowledge codification, which requires active and intentional sense-making processes and a significant cognitive investment, generates a deep understanding of past experience and therefore diminishes the likelihood of inappropriate generalizations (Finkelstein and Haleblian, 2002; Zollo and Winter, 2002). In turn, knowledge codification should bring greater organizational adaptation and therefore reduce persistence in decision-making. Another stream of research has noted that the tools produced through knowledge codification may also act as a source of behavioral control that specifies how decision makers should act in a given situation based on prior experience (Hoskisson and Hitt, 1988; Snell, 1992). Therefore, the codification of knowledge in written tools may also reinforce inertial behavior. We reconcile these perspectives and argue that at low levels of codification, the “behavioral control” effect dominates, whereas when codification rises above a certain threshold, deeper learning of the cause-and-effect relationships takes place. We test these theoretical arguments by focusing on a strategic event common to most large organizations—mergers and acquisitions (M&As)—and more specifically on the decision of whether to integrate the target firm after acquiring it. In particular, we study how experience and knowledge codification regarding the integration decision affect the probability of integrating a newly acquired target firm. To this end, we use a sample of US bank mergers realized between 1985 and 1995. Although these data are relatively old, they are still appropriate for testing our theory for three reasons. First, the US commercial banking industry engaged in several acquisitions after 1985, and therefore, the conditions were ripe for the accumulation of experience and the codification of knowledge. Second, since 1996, US banks have pursued strategies (Hagendorff and Keasey, 2009; Hannan and Pilloff, 2009) that are likely to more consistently result in target integration, reducing the likelihood of observing variation in post-acquisition integration decisions. Third, the sample contains reliable (and typically rare) data on integration decisions and knowledge codification. Nevertheless, we believe that the process studied here is not specific to a given business cycle or industry and that our data, though not recent, are subject to the usual generalizability caveats of any empirical paper. Consistent with previous literature, our findings show that experience with the integration decision increases the likelihood of integrating the target firm in the following acquisition. Moreover, we find evidence suggesting that persistence in decision-making is less likely to be due to superior capabilities in the integration process than to inertia, which is consistent with the view that “the stability-providing effect of routines does […] develop into a pathology” (Becker, 2004: 659). Most importantly, we show that whereas knowledge codification initially increases persistence in decision-making, the effect reverses at high levels of codification, and codification diminishes the effect of prior experience, thus helping organizations to overcome inertia. These results offer two main contributions to the literature. First, on a general level, we advance the organizational learning literature by shedding new light on several contradictory claims and unclear results on the impact of knowledge codification, which has been characterized as a source of either inertia (i.e., a coercive factor) or learning (i.e., an enabling factor) (Adler and Borys, 1996). We reveal that the role of knowledge codification changes from coercive to enabling as knowledge codification increases. Therefore, knowledge codification could be a remedy to organizational inertia, but only after codification rises above a certain threshold. Second, on a more applied level, we advance the M&A literature by providing new evidence on the determinants of integration decisions that—although crucial for explaining acquisition success—have received relatively scant attention in prior research. 2. Theory and hypotheses 2.1 Integration experience and persistence in decision-making A number of empirical studies have shown that persistence in decision-making pervades organizational life, characterizing various decisions related to a firm’s corporate strategy. For example, Haleblian et al. (2006) find that as an acquirer’s acquisition experience increases, its likelihood to make a subsequent acquisition increases. Bergh and Lim (2008) document that as firms gain experience with sell-offs, they continue to use sell-offs as a preferred corporate restructuring strategy. Additionally, Gulati and Gargiulo (1999) show that previous ties between organizations increase the probability of a future alliance between them, while Baum et al. (2000) find that Canadian nursing home chains tend to acquire targets similar to their prior acquisitions. Overall, several findings in this stream of research provide consistent evidence that prior experience is an important predictor of future behaviors and that experience leads to persistence in strategic decision-making. The core insight is that this process occurs because of the emergence of routines (Nelson and Winter, 1982), which are of fundamental importance to understanding organizational change (Becker et al., 2005). Gavetti and Levinthal (2000) note that “routines reflect experiential wisdom in that they are the outcome of trial and error learning and the selection and retention of prior behaviors” (p. 113). Similarly, Pentland and Feldman (2005) contend that because “any particular organizational routine can exhibit a great deal of continuity over time” (p. 794), routines are often considered a source of organizational stability (see Becker, 2004 for a review). Persistence in decision-making might also affect the empirical context of this study, that is, post-acquisition integration decisions. Gavetti and Rivkin (2007) provide an interesting example in this respect. In 1997, the Internet portal Lycos completed its first acquisition, Tripod—a homepage site giving users software tools and server space to build their own Web pages. After much debate on how to handle Tripod’s acquisition, “the management team decided to keep the Tripod brand and integrate Tripod selectively … Lycos repeated this decision to maintain multiple brands during the subsequent years as it acquired other sites such as Guestworld, WhoWhere?, Internet Music Distribution, etc. in rapid succession” (Gavetti and Rivkin, 2007: 426). While research on organizational learning offers ample evidence that as an organization becomes more experienced with a particular decision it becomes more likely to repeat it, it remains unclear whether this persistence is caused by learning or by inertia. Persistence in decision-making can contribute to the formation of superior organizational capabilities and could therefore positively impact performance. For example, take a serial acquirer that has made several acquisitions in its history and has integrated the vast majority of the acquired targets. Through accumulated integration experience, this serial acquirer could develop integration routines that help it handle the complex post-merger integration process. Routines’ stability-providing effect is important for learning because this “stability provides a base against which to assess change, compare and learn” (Becker, 2004: 659). Moreover, stability in organizational routines leads to predictability, which in turn aids coordination (Cyert and March, 1963; Nelson and Winter, 1982). Therefore, as noted by Haleblian et al. (2006: 358), “routines can become a source of competitive advantage” and generate persistence in decision-making based on learning and superior capabilities. However, while routines are likely to generate learning in the context of simple and operational organizational tasks, their role in complex and strategic decisions is less clear and can actually become harmful (Zollo, 2009). As noted by Becker (2004: 659), “At times, the stability-providing effect of routines does, however, develop into a pathology.” Routines can contribute to the formation of fixed and inflexible organizational behaviors, leading an organization to develop a lock-in and inertial decision process (Sydow et al., 2009). Whereas persistence in decision-making should not be conceived as a state of total organizational rigidity, it de facto reduces the variance in possible behaviors and decisions, which can be particularly problematic when the decision at hand is complex, such as the post-acquisition integration decision. Based on this understanding and consistent with prior related literature, we predict that persistence in decision-making could be due to inertia and propose the following hypothesis: H1. Due to inertia, as an acquirer’s experience in integration increases, its probability of integrating a newly acquired target increases. 2.2 Integration experience and knowledge codification Although persistence in decision-making pervades organizational life and, as we argue, is likely due to inertia rather than to learning, to date, the literature provides only a limited understanding of the factors that may reduce it. Thus, in this study, we examine how knowledge codification affects the persistence in decision-making caused by the accumulation of integration experience. Knowledge codification refers to the codification of organizational “understandings of the performance implications of internal routines in written tools, such as manuals, blueprints, spreadsheets, decision support systems, project management software, etc.” (Zollo and Winter, 2002: 342). Once knowledge is codified, less tacit knowledge remains “idiosyncratic to a person or few people, and more of it is transformed into some systematic form that can be communicated” (Cowan and Foray, 1997: 595) over time and space at a low marginal cost (Foray and Steinmueller, 2003; Prencipe and Tell, 2001). Moreover, different types of knowledge—from the less complex to the more complex—can be contained in these tools: (i) know-what, or the content, the information, and the facts; (ii) know-how, or the methodology and the procedure; and (iii) know-why, or the rationale, principles, theories, and causalities (Foray and Steinmueller, 2003; Håkanson, 2007; Kale and Singh, 2007). Knowledge codification can influence decision-making persistence because of its two outputs. The first and simplest output of knowledge codification is the provision of ready-to-use solutions. Codified knowledge “is inscripted in a memorization medium, usually a document” (Balconi et al., 2007: 833) that “serves inter alia as a storage depository, as a reference point” (Cowan et al., 2000: 169). These codified tools make it easier for organizations to address recurring problems that present themselves over time with similar characteristics, and this effect is readily observable. The second output of knowledge codification is realized over time. The process of creating and updating codification tools implies an effort to truly understand the output of past actions, even when learning is not the deliberate goal (Zollo and Winter, 2002). Therefore, knowledge codification goes well beyond the output of ready-to-use solutions and can provide a more profound understanding of the criteria to consider when making complex decisions. In other words, decision makers—by involving themselves in codification effort—emerge with a crisper understanding of the causal linkages between decisions and outcomes, that is, what works and what fails, under what conditions, and—most importantly—why (Kale and Singh, 2007). The reason is that decision makers, by repeatedly undergoing codification efforts, can identify the strengths and weaknesses of existing codified tools and change them accordingly (Gavetti and Levinthal, 2000). Moreover, knowledge codification presupposes knowledge articulation (Håkanson, 2007), or “the process through which implicit knowledge is articulated through collective discussions, debriefing sessions, and performance evaluation processes” (Zollo and Winter, 2002: 341). Therefore, the process of knowledge codification produces learning that goes above and beyond the tools in which knowledge is codified. This effect, however, develops over time, and only after sufficient effort. These two outputs of knowledge codification can exert different, contrasting moderating effects on the positive relationship between experience and decision-making persistence (i.e., H1). On one hand, the first output—the production of ready-to-use solutions captured in documents—may reinforce the inertial effect of experience. Hoskisson and Hitt (1988) and Snell (1992) contend that codified tools represent a form of behavioral control that dictates how decision makers should act in a given situation. Codified tools are repositories of organizational memory (Cyert and March, 1963; Nonaka, 1994; Zander and Kogut, 1995) that are designed to ensure that everyone complies with established organizational practices. In this respect, the knowledge recorded in these tools can serve not only as a reference point in the decision-making process but also “possibly as an authority” (Cowan et al., 2000), even in cases different from those that led to codification. It has been argued that codified tools can do much more than provide a simple possible ready-to-use solution to problems the firm encounters; they could even prescribe “who should do what, when and under which conditions” (Schulz, 1998: 847), limiting decision makers’ autonomy. Therefore, organizational tools enable the firm to address recurring problems in a pre-programmed way, leading to inertia in terms of the application of predefined solutions and increasing organizational inertia. In this respect, for instance, Benner and Tushman (2002) find that ISO 9000 quality program certifications—which include a vast number of written organizational procedures—favor the selection of incremental innovations that are similar to what a firm has done and is doing at the expense of exploratory ones. In our empirical context, knowledge codification could thus reinforce the pattern set by past decisional persistence. If the tools resulting from knowledge codification function as behavioral-control mechanisms, then knowledge codification should strengthen the impact of past integration experience on the probability of integrating a newly acquired target. In contrast, the learning process that occurs over time through codification should have the opposite impact. Given that knowledge codification produces a crisper understanding of the causal linkages between decisions and outcomes, then it should weaken the impact of past decisional persistence on the probability of integrating a newly acquired target. We propose that these two opposing effects can be reconciled and provide a non-monotonic moderating effect of knowledge codification because they emerge at different speeds, and therefore, one effect can prevail over the other at different levels of codification. In particular, because it emerges immediately, the behavioral-control effect of knowledge codification may trump the positive role of learning and allow knowledge codification to increase the persistence in decision-making caused by prior experience—rather than reduce it—until sufficient levels of knowledge codification accumulate. At this point, the effect of learning prevails. Specifically, applying this logic to our empirical context implies that knowledge codification reinforces the inertial effect of integration experience at low levels of codification but reverses it at high levels. The tools that result from low levels of knowledge codification could induce a firm to persist along its decision pattern, as they function as behavioral-control mechanisms before they begin producing a deep understanding of cause-and-effect relationships. In contrast, as the level of codification rises above a certain threshold, the firm becomes able to handle the focal acquisition in a more flexible way, which allows for decision variation according to the specificities of each deal. Formally, we hypothesize the following: H2. As an acquiring organization increases its level of knowledge codification related to integration, the positive impact of the integration experience on the probability of integrating a newly acquired target strengthens and then weakens. Note that whenever both positive and negative effects are at play, the relationship’s precise shape is fundamentally an empirical issue (see Haans et al., 2016 for a review). For instance, some arguments seem to suggest that the inertial effect generated by increasing levels of knowledge codification could prevail over the positive effects of learning. In our context, we expect that the positive effects of codified tools will prevail at high levels of knowledge codification. A high number of codified tools—which can strongly prescribe what to do in the context of simple and repetitive operational activities—cannot precisely and stringently prescribe what to do in our context. Post-acquisition integration decisions are complex, infrequent strategic decisions that cannot be taken mindlessly and automatically based on codified tools alone (Zollo, 2009; Castellaneta and Salvato, 2017). Therefore, higher levels of codified tools should sustain the development of organizational learning without reducing the variance in the observed decision. Thus, a priori, we expect an inverted U-shaped moderation of knowledge codification. 3. Data 3.1 Sample To test our hypotheses, we investigate M&As in the US commercial banking industry between 1985 and 1995.1 Although these data are relatively old, this setting and period offer several benefits. First, the US commercial banking industry started to become very active during that decade, so the conditions were ripe for experience accumulation and knowledge codification and for the emergence of their (positive or negative) effects. Second, evidence suggests that banks displayed significant differences in how they made post-acquisition integration decisions during this period. From 1996 on, however (see DeYoung et al., 2009, for a review), US banks began to pursue mainly efficiency strategies by acquiring inefficient counterparts (Hannan and Pilloff, 2009) or synergy strategies by sharing resources to achieve common objectives (Hagendorff and Keasey, 2009). Both strategies are likely to result in post-acquisition integration and therefore to reduce the likelihood of observing variation in this post-acquisition integration decision. Third, and arguably most importantly, our data set provides rich and fine-grained information on two variables of interest that are rarely available in public data sets and that relate to integration decisions and to the extent of knowledge codification. The research design involved three phases. First, fieldwork was conducted in 12 banks that were active acquirers. This fieldwork sheds light on acquisition practices in the commercial banking industry. Second, 45 decision makers were interviewed to develop a questionnaire-based survey and to check for its measurability and clarity. Third, the survey was administrated among the 250 largest bank-holding companies in the United States in 1996. The survey had two main parts: a profile of acquisition history and a questionnaire for the acquiring bank. In the first portion, respondents were asked to list all acquisitions conducted by their bank and to give basic information about each acquisition, such as asset sizes, the degree of market relatedness, pre-acquisition profitability, and the level of integration. The acquiring-bank questionnaire also elicited information about the ad hoc tools that were developed to manage the acquisition process (e.g., integration manuals, systems conversion manuals, training packages, due-diligence checklists, branch staffing models, and product mapping models). Of the 250 bank-holding companies contacted, 70 did not undertake an acquisition after 1985, and 16 were acquired during the invitation period. Of the remaining 164 banks, 51 responded to the survey, resulting in a 31.7% response rate. To check whether the respondent banks were different from the contacted ones, a standard means comparison test was performed. Responding organizations did not differ from the original set of 250 bank-holding companies in terms of their return on assets (ROA), equity, or efficiency ratios, although they tended to be larger in terms of total assets (P < 0.05). The response rate also appeared to be satisfactory, given the seniority of the respondents and the complexity of the survey. This response rate could also have resulted due to the salience of this topic to industry participants and the in-depth pretests of the survey instrument (Groves et al., 1992). The respondents, who were the most knowledgeable persons at each bank, were identified through telephone calls prior to the mailing and included managers responsible for corporate development or the M&A group (25 cases), coordinators of post-acquisition integration processes (14 cases), CFOs (9 cases), and CEOs (3 cases). Completing the questionnaire offered these respondents an opportunity to benchmark their acquisition practices with those of other firms in the industry. Moreover, the respondents received assurances that their individual responses would remain confidential. Four bank-holding companies were excluded from the analysis because of incomplete responses, and 16 more were excluded because of missing values. The final sample contains data on 31 acquiring banks that completed 408 acquisitions during the period under observation. Due to missing values for same variables, different subsamples are used depending on the variables included in the model. 3.2 Dependent variable 3.2.1 Integration At the core of our study lies the critical strategic decision on the integration of an acquired target’s structures. To measure whether a target was integrated, we use a dummy variable that was equal to 1 if “all systems, procedures, and products (of the two firms) were completely integrated” and equal to 0 in the remaining three cases (if “many but not all systems, procedures, and products were aligned or centralized”; “if only selected systems, procedures, or products were aligned or centralized”; or if “few or no features were aligned or centralized”).2 This measure considers the fact that when a target is not fully integrated, the parent has two options: (i) to centralize some of the target’s activities or (ii) to simply align the activities between the parent and the target, ensuring that these activities are run similarly in both companies. 3.3. Explanatory and control variables 3.3.1 Integration experience Consistent with the previous literature, we hypothesized that as an organization becomes more experienced with a particular decision, it becomes more likely to repeat that decision (Amburgey et al., 1993; Gulati, 1995; Shaver et al., 1997; Dimov et al., 2012). As we apply this theory in the M&A context and, more precisely, considering post-M&A integration decisions, we measure (past) decisional persistence as (the log of) the number of past acquisitions that were integrated. This measure captures the experience that a firm has with a particular decision. 3.3.2 Knowledge codification The measure of knowledge codification is based on the number of tools developed by the acquiring firm at the time of the focal acquisition. These tools reflect the various parts of the acquisition process, such as financial evaluation, due diligence, conversion of information systems, human resource integration, and sales/product integration. The number of tools ranges from 1 to 11 and refers to the due-diligence checklist and to separate manuals on: financial evaluation spreadsheets, due diligence checklist, due diligence manual, info systems conversion manual, info systems training manual, affiliation/integration manual, branch staffing models, training/self-training packages, products training manual, product mapping models, and project management packages (Zollo and Singh, 2004; Zollo, 2009). Consistent with the feedback received in the fieldwork, this list also contains tools—such as the due diligence checklist—that, while not directly linked to the integration process, help in its planning. This measure thereby approximates the acquiring bank’s development of integration practices at the time of the focal acquisition. As with integration experience, we used a log transformation.3 3.3.3 Control variables First, we control for the relatedness between the acquiring firm and the target. Higher market relatedness between the acquiring firm and the target might increase the likelihood of integration because more functions could overlap in their extension and routines. The bank industry clearly distinguishes between “in-market” (horizontal) and “out-market” (market extension) acquisitions; the survey probes the categorization of each transaction listed in the acquisition history profile. Thus, resource relatedness is coded as 1 if the acquisition is in-market and 0 if it is out-market. Second, we control for target quality. The quality of the acquired firm can affect the decision to integrate it such that acquiring firms are less likely to integrate a high-quality target. To measure the pre-acquisition quality of the acquired firm’s resource endowments, we use the following scale: −2 (the acquired institution was bankrupt), −1 (it was a poor performer), 0 (it was an average performer), +1 (it was a good performer), and +2 (it was an outstanding performer). Third, we control for relative acquisition size. The acquired firm’s size relative to that of the acquiring bank (based on total assets) can be stated as a percentage (Datta, 1991). Fourth, we control for parallel acquisitions, that is, the number of acquisitions the acquiring firm manages simultaneously might influence post-acquisition strategies, including integration. A firm that has carried out several acquisitions might lack the energy, time, or resources (Castellaneta and Zollo, 2015) to pursue deep organizational integration. However, more acquisitions might signal the firm’s rapid growth strategy, which can be attained only through strict integration. We account for this effect by controlling for the number of acquisitions concluded by the acquiring firm in the same year as the focal acquisition. Finally, we also control for acquisition experience, computed as the number of acquisitions completed by the acquiring firm before the focal acquisitions (irrespective of whether they were integrated).4 Note that acquisition experience and codification do not necessarily covary: some firms carry out several acquisitions without conducting any knowledge codification, while other firms begin knowledge codification at a very early stage in their acquisition pattern. 4. Analyses and results We study the effect of past decisional experience, knowledge codification, and their interaction on the probability of integrating a target after an M&A deal. Given the binary nature of the dependent variable, we use as the main model a logistic regression with robust standard errors, which accounts for heteroscedasticity in the error distribution. Table 1 reports descriptive statistics and a correlation matrix for the variables used in our hypothesis tests. Table 2 presents the results of the logit model, where explanatory variables are inserted gradually. Table 1. Descriptive statistics and correlation matrix Variables Mean SD 1 2 3 4 5 6 7 8 9 1. Integration 0.718 0.450 1 2. Integration experience 5.027 6.832 0.378*** 1 3. Acquisition experience 11.476 10.162 0.135** 0.637*** 1 4. Relatedness 0.623 0.485 0.417*** 0.221*** 0.168*** 1 5. Parallel acquisitions 3.630 2.826 0.205*** 0.658*** 0.507*** 0.135** 1 6. Target quality −0.0234 1.076 −0.260*** −0.0454 0.0349 −0.223*** 0.0537 1 7. Relative acquisition size 1.057 1.665 −0.0385 0.0455 0.0565 −0.171*** −0.0577 0.0820 1 8. ROA 0.007 0.376 0.239*** 0.128* 0.0115 0.0961 0.195*** −0.120* 0.0789 1 9. Deviation 0.281 0.251 −0.606*** −0.424*** −0.267*** −0.459*** −0.435*** 0.252*** 0.124* −0.263*** 1 10. PoIAs 0.432 0.390 0.516*** 0.612*** 0.0193 0.165*** 0.172*** −0.0901 0.0824 0.135* −0.256*** 1 11. Knowledge codification 4.854 3.677 0.0396 0.328*** 0.459*** 0.0272 0.367*** 0.179*** 0.0766 0.0998 −0.141** 0.0595 Variables Mean SD 1 2 3 4 5 6 7 8 9 1. Integration 0.718 0.450 1 2. Integration experience 5.027 6.832 0.378*** 1 3. Acquisition experience 11.476 10.162 0.135** 0.637*** 1 4. Relatedness 0.623 0.485 0.417*** 0.221*** 0.168*** 1 5. Parallel acquisitions 3.630 2.826 0.205*** 0.658*** 0.507*** 0.135** 1 6. Target quality −0.0234 1.076 −0.260*** −0.0454 0.0349 −0.223*** 0.0537 1 7. Relative acquisition size 1.057 1.665 −0.0385 0.0455 0.0565 −0.171*** −0.0577 0.0820 1 8. ROA 0.007 0.376 0.239*** 0.128* 0.0115 0.0961 0.195*** −0.120* 0.0789 1 9. Deviation 0.281 0.251 −0.606*** −0.424*** −0.267*** −0.459*** −0.435*** 0.252*** 0.124* −0.263*** 1 10. PoIAs 0.432 0.390 0.516*** 0.612*** 0.0193 0.165*** 0.172*** −0.0901 0.0824 0.135* −0.256*** 1 11. Knowledge codification 4.854 3.677 0.0396 0.328*** 0.459*** 0.0272 0.367*** 0.179*** 0.0766 0.0998 −0.141** 0.0595 Note: * P < 0.05; ** P < 0.01; *** P < 0.001. Table 1. Descriptive statistics and correlation matrix Variables Mean SD 1 2 3 4 5 6 7 8 9 1. Integration 0.718 0.450 1 2. Integration experience 5.027 6.832 0.378*** 1 3. Acquisition experience 11.476 10.162 0.135** 0.637*** 1 4. Relatedness 0.623 0.485 0.417*** 0.221*** 0.168*** 1 5. Parallel acquisitions 3.630 2.826 0.205*** 0.658*** 0.507*** 0.135** 1 6. Target quality −0.0234 1.076 −0.260*** −0.0454 0.0349 −0.223*** 0.0537 1 7. Relative acquisition size 1.057 1.665 −0.0385 0.0455 0.0565 −0.171*** −0.0577 0.0820 1 8. ROA 0.007 0.376 0.239*** 0.128* 0.0115 0.0961 0.195*** −0.120* 0.0789 1 9. Deviation 0.281 0.251 −0.606*** −0.424*** −0.267*** −0.459*** −0.435*** 0.252*** 0.124* −0.263*** 1 10. PoIAs 0.432 0.390 0.516*** 0.612*** 0.0193 0.165*** 0.172*** −0.0901 0.0824 0.135* −0.256*** 1 11. Knowledge codification 4.854 3.677 0.0396 0.328*** 0.459*** 0.0272 0.367*** 0.179*** 0.0766 0.0998 −0.141** 0.0595 Variables Mean SD 1 2 3 4 5 6 7 8 9 1. Integration 0.718 0.450 1 2. Integration experience 5.027 6.832 0.378*** 1 3. Acquisition experience 11.476 10.162 0.135** 0.637*** 1 4. Relatedness 0.623 0.485 0.417*** 0.221*** 0.168*** 1 5. Parallel acquisitions 3.630 2.826 0.205*** 0.658*** 0.507*** 0.135** 1 6. Target quality −0.0234 1.076 −0.260*** −0.0454 0.0349 −0.223*** 0.0537 1 7. Relative acquisition size 1.057 1.665 −0.0385 0.0455 0.0565 −0.171*** −0.0577 0.0820 1 8. ROA 0.007 0.376 0.239*** 0.128* 0.0115 0.0961 0.195*** −0.120* 0.0789 1 9. Deviation 0.281 0.251 −0.606*** −0.424*** −0.267*** −0.459*** −0.435*** 0.252*** 0.124* −0.263*** 1 10. PoIAs 0.432 0.390 0.516*** 0.612*** 0.0193 0.165*** 0.172*** −0.0901 0.0824 0.135* −0.256*** 1 11. Knowledge codification 4.854 3.677 0.0396 0.328*** 0.459*** 0.0272 0.367*** 0.179*** 0.0766 0.0998 −0.141** 0.0595 Note: * P < 0.05; ** P < 0.01; *** P < 0.001. Table 2. Logit predicting integration probability with integration experience Variables (1) (2) (3) (4) (5) (6) Integration Integration Integration Integration Integration Integration Integration experience 0.952*** 1.115*** 1.106*** 0.962*** 1.188*** (0.150) (0.170) (0.171) (0.295) (0.277) Knowledge codification 0.209 0.388 0.134 −0.079 (0.166) (0.264) (0.188) (0.328) Integration experience × knowledge codification 0.105 0.670** (0.167) (0.273) Knowledge codification ^2 −0.158 0.085 (0.177) (0.199) Integration experience × knowledge codification ^2 −0.330** (0.137) Acquisition experience −0.007 −0.047** −0.041 −0.034 −0.039 −0.036 (0.015) (0.022) (0.025) (0.026) (0.026) (0.026) Relatedness 1.941*** 2.140*** 2.378*** 2.398*** 2.363*** 2.545*** (0.291) (0.355) (0.396) (0.398) (0.396) (0.412) Parallel acquisitions 0.391*** 0.369*** 0.210** 0.236** 0.208** 0.255** (0.077) (0.083) (0.095) (0.100) (0.097) (0.103) Target quality −0.459*** −0.358* −0.484** −0.470** −0.488** −0.532*** (0.138) (0.185) (0.206) (0.201) (0.204) (0.204) Relative acquisition size 0.069 0.021 −0.109 −0.103 −0.119 −0.112 (0.070) (0.071) (0.095) (0.097) (0.099) (0.102) Constant −1.254*** −1.574*** −1.612*** −1.564*** −1.503*** −1.716*** (0.344) (0.400) (0.445) (0.465) (0.487) (0.531) Observations 384 384 358 358 358 358 Pseudo R-squared 0.255 0.398 0.451 0.453 0.452 0.470 Log likelihood −161.592 −130.664 −114.345 −113.950 −114.066 −110.317 Variables (1) (2) (3) (4) (5) (6) Integration Integration Integration Integration Integration Integration Integration experience 0.952*** 1.115*** 1.106*** 0.962*** 1.188*** (0.150) (0.170) (0.171) (0.295) (0.277) Knowledge codification 0.209 0.388 0.134 −0.079 (0.166) (0.264) (0.188) (0.328) Integration experience × knowledge codification 0.105 0.670** (0.167) (0.273) Knowledge codification ^2 −0.158 0.085 (0.177) (0.199) Integration experience × knowledge codification ^2 −0.330** (0.137) Acquisition experience −0.007 −0.047** −0.041 −0.034 −0.039 −0.036 (0.015) (0.022) (0.025) (0.026) (0.026) (0.026) Relatedness 1.941*** 2.140*** 2.378*** 2.398*** 2.363*** 2.545*** (0.291) (0.355) (0.396) (0.398) (0.396) (0.412) Parallel acquisitions 0.391*** 0.369*** 0.210** 0.236** 0.208** 0.255** (0.077) (0.083) (0.095) (0.100) (0.097) (0.103) Target quality −0.459*** −0.358* −0.484** −0.470** −0.488** −0.532*** (0.138) (0.185) (0.206) (0.201) (0.204) (0.204) Relative acquisition size 0.069 0.021 −0.109 −0.103 −0.119 −0.112 (0.070) (0.071) (0.095) (0.097) (0.099) (0.102) Constant −1.254*** −1.574*** −1.612*** −1.564*** −1.503*** −1.716*** (0.344) (0.400) (0.445) (0.465) (0.487) (0.531) Observations 384 384 358 358 358 358 Pseudo R-squared 0.255 0.398 0.451 0.453 0.452 0.470 Log likelihood −161.592 −130.664 −114.345 −113.950 −114.066 −110.317 Note: Robust standard errors in parentheses. * P < 0.1; ** P < 0.05; *** P < 0.01. Table 2. Logit predicting integration probability with integration experience Variables (1) (2) (3) (4) (5) (6) Integration Integration Integration Integration Integration Integration Integration experience 0.952*** 1.115*** 1.106*** 0.962*** 1.188*** (0.150) (0.170) (0.171) (0.295) (0.277) Knowledge codification 0.209 0.388 0.134 −0.079 (0.166) (0.264) (0.188) (0.328) Integration experience × knowledge codification 0.105 0.670** (0.167) (0.273) Knowledge codification ^2 −0.158 0.085 (0.177) (0.199) Integration experience × knowledge codification ^2 −0.330** (0.137) Acquisition experience −0.007 −0.047** −0.041 −0.034 −0.039 −0.036 (0.015) (0.022) (0.025) (0.026) (0.026) (0.026) Relatedness 1.941*** 2.140*** 2.378*** 2.398*** 2.363*** 2.545*** (0.291) (0.355) (0.396) (0.398) (0.396) (0.412) Parallel acquisitions 0.391*** 0.369*** 0.210** 0.236** 0.208** 0.255** (0.077) (0.083) (0.095) (0.100) (0.097) (0.103) Target quality −0.459*** −0.358* −0.484** −0.470** −0.488** −0.532*** (0.138) (0.185) (0.206) (0.201) (0.204) (0.204) Relative acquisition size 0.069 0.021 −0.109 −0.103 −0.119 −0.112 (0.070) (0.071) (0.095) (0.097) (0.099) (0.102) Constant −1.254*** −1.574*** −1.612*** −1.564*** −1.503*** −1.716*** (0.344) (0.400) (0.445) (0.465) (0.487) (0.531) Observations 384 384 358 358 358 358 Pseudo R-squared 0.255 0.398 0.451 0.453 0.452 0.470 Log likelihood −161.592 −130.664 −114.345 −113.950 −114.066 −110.317 Variables (1) (2) (3) (4) (5) (6) Integration Integration Integration Integration Integration Integration Integration experience 0.952*** 1.115*** 1.106*** 0.962*** 1.188*** (0.150) (0.170) (0.171) (0.295) (0.277) Knowledge codification 0.209 0.388 0.134 −0.079 (0.166) (0.264) (0.188) (0.328) Integration experience × knowledge codification 0.105 0.670** (0.167) (0.273) Knowledge codification ^2 −0.158 0.085 (0.177) (0.199) Integration experience × knowledge codification ^2 −0.330** (0.137) Acquisition experience −0.007 −0.047** −0.041 −0.034 −0.039 −0.036 (0.015) (0.022) (0.025) (0.026) (0.026) (0.026) Relatedness 1.941*** 2.140*** 2.378*** 2.398*** 2.363*** 2.545*** (0.291) (0.355) (0.396) (0.398) (0.396) (0.412) Parallel acquisitions 0.391*** 0.369*** 0.210** 0.236** 0.208** 0.255** (0.077) (0.083) (0.095) (0.100) (0.097) (0.103) Target quality −0.459*** −0.358* −0.484** −0.470** −0.488** −0.532*** (0.138) (0.185) (0.206) (0.201) (0.204) (0.204) Relative acquisition size 0.069 0.021 −0.109 −0.103 −0.119 −0.112 (0.070) (0.071) (0.095) (0.097) (0.099) (0.102) Constant −1.254*** −1.574*** −1.612*** −1.564*** −1.503*** −1.716*** (0.344) (0.400) (0.445) (0.465) (0.487) (0.531) Observations 384 384 358 358 358 358 Pseudo R-squared 0.255 0.398 0.451 0.453 0.452 0.470 Log likelihood −161.592 −130.664 −114.345 −113.950 −114.066 −110.317 Note: Robust standard errors in parentheses. * P < 0.1; ** P < 0.05; *** P < 0.01. H1 posited that as integration experience increases, the probability of integrating a newly acquired target increases. The positive and significant coefficient of integration experience (Model 6 of Table 2) provides support for H1 (P < 0.01). However, we also claimed that integration experience likely produces persistence in decision-making not because of learning but rather because of inertia. However, the fact that a firm repeats the same decision over time does not automatically imply that the decision is harmful for the organization. In contrast, one might argue that a decision is repeated over time because of increasing returns created by the feedback between experience and competence. Hence, the acquiring firm might continue to conduct the same “type of deal” for which it is advisable to continue integrating because of the firm’s knowledge of and capabilities with respect to the integration process. If this is the case, then there is no need to diminish persistence in decision-making. Because we cannot directly observe whether superior capabilities or inertia drive the persistence effect, we resort to preliminary indirect empirical evidence according to the following steps. Step 1: Predicting the integration decision. In the first step, we predict—through a logit model—the expected integration decision for each deal (i.e., the probability, pint, that two merging firms are actually integrated) based on four relevant predictors of the integration decision that are consistently suggested by the related literature. In other words, we first estimate a logit model in which the dependent variable is the integration decision and the covariates are relatedness, target quality, relative acquisition size, and parallel acquisitions. Given the result of this estimation, for each deal, we then predict the probability pint that the target is integrated, conditional on the specific values that the four covariates assume in the specific deal. In essence, the pint variable thus captures how an acquiring firm in our sample would handle, on average, its post-acquisition integration decision, given the actual value of the four predictors. By construction, pint is bounded by 0 and 1. Step 2: Calculating the deviation. In the second step, for each deal, we create a Deviation variable such that Deviation = | Actual integration decision – pint |. If the predicted pint is low (i.e., the four variables suggest that the deal was unlikely to be integrated) but the target was eventually integrated, the value of Deviation is relatively high. If the deal was integrated and pint is high, then the value of Deviation is low. Similarly, if the deal was not actually integrated and pint is low, Deviation is negative but has a low absolute value, whereas if pint is high, Deviation is negative and has a high absolute value. Step 3: Deviation and performance. In the third step, we calculate a measure of deal performance and regress it on the absolute value of Deviation and the control variables used in the first step. Following Zollo and Singh (2004), we measure acquisition performance as the difference between the ROA of the acquiring bank 3 years after the acquisition (which considers the entire acquiring bank, including the acquired target) versus 1 year before (which considers only the acquiring bank).5 This step checks whether a decision regarding organizational integration—that “deviates” from what deal characteristics indicate an average sample firm would do—has a positive or a negative effect on performance. As the previous regressions have highlighted, this possible “deviation” is due to persistence in decision-making.6 However, persistence due to increasing returns and superior capabilities would be mirrored in a positive effect of deviation on performance. Conversely, if persistence were related to organizational inertia, we would expect the effect of deviation to be negative. The results of this last linear regression, which is reported in Model 2 of Table 3, show that the parameter estimate of the absolute value of Deviation is negative and significant (P < 0.1). Therefore, with all other things being equal, as a firm deviates more from the “predicted” decision (i.e., deviates from what the firms in our sample would have done on average in that specific situation), its deal performance decreases. Hence, this result suggests that persistence in decision-making is associated with negative results in terms of performance. Clearly, these results are only indicative and correlational in nature, as our data structure does not allow us to properly disentangle inertia from learning; still, for this sample, preliminary evidence is consistent with the notion that persistence and organizational inertia are associated. Table 3. The effect of relatedness, parallel acquisitions, target quality, and relative acquisition size on integration probability (Model 1); the effect of deviation (absolute value) on acquisition performance measured as ROA (Model 2) Variables (1) (2) Integration Acquisition performance (ROA) Deviation (absolute value) −0.275* (0.117) Relatedness 1.912*** 0.029 (0.276) (0.058) Parallel acquisitions 0.375*** 0.015** (0.071) (0.005) Target quality −0.456*** −0.030 (0.133) (0.024) Relative acquisition size 0.078 0.025 (0.068) (0.015) Constant −1.275*** −0.043 (0.324) (0.072) Observations 408 259 Log likelihood −175.332 0.095 Variables (1) (2) Integration Acquisition performance (ROA) Deviation (absolute value) −0.275* (0.117) Relatedness 1.912*** 0.029 (0.276) (0.058) Parallel acquisitions 0.375*** 0.015** (0.071) (0.005) Target quality −0.456*** −0.030 (0.133) (0.024) Relative acquisition size 0.078 0.025 (0.068) (0.015) Constant −1.275*** −0.043 (0.324) (0.072) Observations 408 259 Log likelihood −175.332 0.095 Note: Robust standard errors in parentheses. * P < 0.1; ** P < 0.05; *** P < 0.01. Table 3. The effect of relatedness, parallel acquisitions, target quality, and relative acquisition size on integration probability (Model 1); the effect of deviation (absolute value) on acquisition performance measured as ROA (Model 2) Variables (1) (2) Integration Acquisition performance (ROA) Deviation (absolute value) −0.275* (0.117) Relatedness 1.912*** 0.029 (0.276) (0.058) Parallel acquisitions 0.375*** 0.015** (0.071) (0.005) Target quality −0.456*** −0.030 (0.133) (0.024) Relative acquisition size 0.078 0.025 (0.068) (0.015) Constant −1.275*** −0.043 (0.324) (0.072) Observations 408 259 Log likelihood −175.332 0.095 Variables (1) (2) Integration Acquisition performance (ROA) Deviation (absolute value) −0.275* (0.117) Relatedness 1.912*** 0.029 (0.276) (0.058) Parallel acquisitions 0.375*** 0.015** (0.071) (0.005) Target quality −0.456*** −0.030 (0.133) (0.024) Relative acquisition size 0.078 0.025 (0.068) (0.015) Constant −1.275*** −0.043 (0.324) (0.072) Observations 408 259 Log likelihood −175.332 0.095 Note: Robust standard errors in parentheses. * P < 0.1; ** P < 0.05; *** P < 0.01. H2 suggests that as an acquiring organization increases its level of knowledge codification related to integration, integration experience’s positive impact on the probability of integrating a newly acquired target strengthens and then weakens. That is, we propose a curvilinear moderating effect. The interaction term between integration experience and knowledge codification is positive and significant (P < 0.05), and that between integration experience and knowledge codification squared is negative and significant (P < 0.05) in Model 6 of Table 2. Overall, these findings support H2. The previous literature has proposed that persistence is a result of the amount of accumulated experience (Amburgey and Miner, 1992). The core idea is that frequent repetitions of a decision cause decision-making persistence and path dependence (Katz and Shapiro, 1985; Levinthal and March, 1993). However, in principle, path dependence can be created at any level of experience. The only condition necessary for the emergence of path-dependent behaviors is that the organization “fall into the trap” of certain constraints, such as routines, cognitive schemata, or structures, which, despite being influenced by repetition, can emerge regardless of the quantity of experience accumulated. Hence, as a robustness check, we operationalize past decision persistence as the percentage of integrated acquisitions (PoIAs) (i.e., the share of integrated deals among all the acquiring firm’s previous deals). Model 6 in Table 4 shows that our results are robust to the use of this operationalization of past persistence. Table 4. Logit predicting integration probability with the PoIAs Variables (1) (2) (3) (4) (5) (6) Integration Integration Integration Integration Integration Integration PoIAs 3.467*** 3.960*** 3.953*** 2.413*** 3.805*** (0.520) (0.563) (0.564) (0.714) (0.995) Knowledge codification 0.117 0.426 −0.475* −0.806* (0.183) (0.297) (0.253) (0.471) PoIA × knowledge codification 1.454*** 2.826*** (0.492) (0.860) Knowledge codification ^ 2 −0.255 0.227 (0.174) (0.269) PoIA × knowledge codification ^ 2 −1.107** (0.540) Acquisition experience −0.008 0.012 0.029 0.039* 0.039* 0.039* (0.014) (0.017) (0.019) (0.021) (0.021) (0.021) Relatedness 1.927*** 2.086*** 2.304*** 2.326*** 2.439*** 2.648*** (0.279) (0.355) (0.386) (0.385) (0.402) (0.430) Parallel acquisitions 0.389*** 0.404*** 0.265*** 0.314*** 0.241** 0.283** (0.075) (0.083) (0.102) (0.110) (0.114) (0.119) Target quality −0.452*** −0.496*** −0.636*** −0.610*** −0.612*** −0.631*** (0.133) (0.177) (0.195) (0.190) (0.193) (0.192) Relative acquisition size 0.081 −0.025 −0.140 −0.133 −0.149 −0.130 (0.068) (0.074) (0.117) (0.120) (0.121) (0.123) Constant −1.254*** −2.889*** −3.069*** −3.020*** −2.479*** −2.946*** (0.327) (0.503) (0.549) (0.584) (0.614) (0.691) Observations 408 384 358 358 358 358 Pseudo R-squared 0.254 0.407 0.453 0.458 0.476 0.491 Log likelihood −175.211 −128.607 −113.885 −112.858 −108.968 −106.028 Variables (1) (2) (3) (4) (5) (6) Integration Integration Integration Integration Integration Integration PoIAs 3.467*** 3.960*** 3.953*** 2.413*** 3.805*** (0.520) (0.563) (0.564) (0.714) (0.995) Knowledge codification 0.117 0.426 −0.475* −0.806* (0.183) (0.297) (0.253) (0.471) PoIA × knowledge codification 1.454*** 2.826*** (0.492) (0.860) Knowledge codification ^ 2 −0.255 0.227 (0.174) (0.269) PoIA × knowledge codification ^ 2 −1.107** (0.540) Acquisition experience −0.008 0.012 0.029 0.039* 0.039* 0.039* (0.014) (0.017) (0.019) (0.021) (0.021) (0.021) Relatedness 1.927*** 2.086*** 2.304*** 2.326*** 2.439*** 2.648*** (0.279) (0.355) (0.386) (0.385) (0.402) (0.430) Parallel acquisitions 0.389*** 0.404*** 0.265*** 0.314*** 0.241** 0.283** (0.075) (0.083) (0.102) (0.110) (0.114) (0.119) Target quality −0.452*** −0.496*** −0.636*** −0.610*** −0.612*** −0.631*** (0.133) (0.177) (0.195) (0.190) (0.193) (0.192) Relative acquisition size 0.081 −0.025 −0.140 −0.133 −0.149 −0.130 (0.068) (0.074) (0.117) (0.120) (0.121) (0.123) Constant −1.254*** −2.889*** −3.069*** −3.020*** −2.479*** −2.946*** (0.327) (0.503) (0.549) (0.584) (0.614) (0.691) Observations 408 384 358 358 358 358 Pseudo R-squared 0.254 0.407 0.453 0.458 0.476 0.491 Log likelihood −175.211 −128.607 −113.885 −112.858 −108.968 −106.028 Note: Robust standard errors in parentheses. * P < 0.1; ** P < 0.05; *** P < 0.01. Table 4. Logit predicting integration probability with the PoIAs Variables (1) (2) (3) (4) (5) (6) Integration Integration Integration Integration Integration Integration PoIAs 3.467*** 3.960*** 3.953*** 2.413*** 3.805*** (0.520) (0.563) (0.564) (0.714) (0.995) Knowledge codification 0.117 0.426 −0.475* −0.806* (0.183) (0.297) (0.253) (0.471) PoIA × knowledge codification 1.454*** 2.826*** (0.492) (0.860) Knowledge codification ^ 2 −0.255 0.227 (0.174) (0.269) PoIA × knowledge codification ^ 2 −1.107** (0.540) Acquisition experience −0.008 0.012 0.029 0.039* 0.039* 0.039* (0.014) (0.017) (0.019) (0.021) (0.021) (0.021) Relatedness 1.927*** 2.086*** 2.304*** 2.326*** 2.439*** 2.648*** (0.279) (0.355) (0.386) (0.385) (0.402) (0.430) Parallel acquisitions 0.389*** 0.404*** 0.265*** 0.314*** 0.241** 0.283** (0.075) (0.083) (0.102) (0.110) (0.114) (0.119) Target quality −0.452*** −0.496*** −0.636*** −0.610*** −0.612*** −0.631*** (0.133) (0.177) (0.195) (0.190) (0.193) (0.192) Relative acquisition size 0.081 −0.025 −0.140 −0.133 −0.149 −0.130 (0.068) (0.074) (0.117) (0.120) (0.121) (0.123) Constant −1.254*** −2.889*** −3.069*** −3.020*** −2.479*** −2.946*** (0.327) (0.503) (0.549) (0.584) (0.614) (0.691) Observations 408 384 358 358 358 358 Pseudo R-squared 0.254 0.407 0.453 0.458 0.476 0.491 Log likelihood −175.211 −128.607 −113.885 −112.858 −108.968 −106.028 Variables (1) (2) (3) (4) (5) (6) Integration Integration Integration Integration Integration Integration PoIAs 3.467*** 3.960*** 3.953*** 2.413*** 3.805*** (0.520) (0.563) (0.564) (0.714) (0.995) Knowledge codification 0.117 0.426 −0.475* −0.806* (0.183) (0.297) (0.253) (0.471) PoIA × knowledge codification 1.454*** 2.826*** (0.492) (0.860) Knowledge codification ^ 2 −0.255 0.227 (0.174) (0.269) PoIA × knowledge codification ^ 2 −1.107** (0.540) Acquisition experience −0.008 0.012 0.029 0.039* 0.039* 0.039* (0.014) (0.017) (0.019) (0.021) (0.021) (0.021) Relatedness 1.927*** 2.086*** 2.304*** 2.326*** 2.439*** 2.648*** (0.279) (0.355) (0.386) (0.385) (0.402) (0.430) Parallel acquisitions 0.389*** 0.404*** 0.265*** 0.314*** 0.241** 0.283** (0.075) (0.083) (0.102) (0.110) (0.114) (0.119) Target quality −0.452*** −0.496*** −0.636*** −0.610*** −0.612*** −0.631*** (0.133) (0.177) (0.195) (0.190) (0.193) (0.192) Relative acquisition size 0.081 −0.025 −0.140 −0.133 −0.149 −0.130 (0.068) (0.074) (0.117) (0.120) (0.121) (0.123) Constant −1.254*** −2.889*** −3.069*** −3.020*** −2.479*** −2.946*** (0.327) (0.503) (0.549) (0.584) (0.614) (0.691) Observations 408 384 358 358 358 358 Pseudo R-squared 0.254 0.407 0.453 0.458 0.476 0.491 Log likelihood −175.211 −128.607 −113.885 −112.858 −108.968 −106.028 Note: Robust standard errors in parentheses. * P < 0.1; ** P < 0.05; *** P < 0.01. The previous results are based on interaction terms in logistic regressions. However, because logit models are nonlinear models, the parameter estimates of the interaction terms should be carefully interpreted. To address this issue, we also adopt a linear probability model approach, which allows us to obtain a direct and simple estimate of the interaction effect. Because the predicted probabilities lie between 0 and 1, the linear probability model does not produce biased estimates (Horrace and Oaxaca, 2006). The results reported in Tables 5 and 6 show that our results are also robust to this specification. Moreover, based on Model 6 in Table 5, we graph the interaction effect between knowledge codification and integration experience (Figure 1). Table 5. Linear probability model predicting integration probability with integration experience Variables (1) (2) (3) (4) (5) (6) Integration Integration Integration Integration Integration Integration Integration experience 0.138*** 0.164*** 0.163*** 0.166*** 0.188*** (0.018) (0.019) (0.019) (0.038) (0.037) Knowledge codification 0.027 0.042 0.028 −0.007 (0.020) (0.034) (0.026) (0.047) Integration experience × Knowledge codification −0.001 0.059 (0.020) (0.036) Knowledge codification ^ 2 −0.012 0.021 (0.021) (0.026) Integration experience × Knowledge codification ^ 2 −0.035** (0.017) Acquisition experience −0.002 −0.009*** −0.007*** −0.007*** −0.007*** −0.008*** (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) Relatedness 0.339*** 0.300*** 0.313*** 0.313*** 0.313*** 0.320*** (0.050) (0.047) (0.049) (0.049) (0.049) (0.049) Parallel acquisitions 0.034*** 0.020*** 0.002 0.003 0.002 0.005 (0.006) (0.006) (0.006) (0.006) (0.007) (0.007) Target quality −0.066*** −0.044** −0.060*** −0.058*** −0.059*** −0.060*** (0.017) (0.018) (0.019) (0.018) (0.019) (0.019) Relative acquisition size 0.012 0.006 −0.008 −0.008 −0.008 −0.008 (0.012) (0.010) (0.014) (0.014) (0.014) (0.014) Constant 0.389*** 0.401*** 0.369*** 0.379*** 0.367*** 0.349*** (0.050) (0.046) (0.051) (0.055) (0.060) (0.063) Observations 384 384 358 358 358 358 R-squared 0.252 0.384 0.452 0.453 0.452 0.462 Variables (1) (2) (3) (4) (5) (6) Integration Integration Integration Integration Integration Integration Integration experience 0.138*** 0.164*** 0.163*** 0.166*** 0.188*** (0.018) (0.019) (0.019) (0.038) (0.037) Knowledge codification 0.027 0.042 0.028 −0.007 (0.020) (0.034) (0.026) (0.047) Integration experience × Knowledge codification −0.001 0.059 (0.020) (0.036) Knowledge codification ^ 2 −0.012 0.021 (0.021) (0.026) Integration experience × Knowledge codification ^ 2 −0.035** (0.017) Acquisition experience −0.002 −0.009*** −0.007*** −0.007*** −0.007*** −0.008*** (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) Relatedness 0.339*** 0.300*** 0.313*** 0.313*** 0.313*** 0.320*** (0.050) (0.047) (0.049) (0.049) (0.049) (0.049) Parallel acquisitions 0.034*** 0.020*** 0.002 0.003 0.002 0.005 (0.006) (0.006) (0.006) (0.006) (0.007) (0.007) Target quality −0.066*** −0.044** −0.060*** −0.058*** −0.059*** −0.060*** (0.017) (0.018) (0.019) (0.018) (0.019) (0.019) Relative acquisition size 0.012 0.006 −0.008 −0.008 −0.008 −0.008 (0.012) (0.010) (0.014) (0.014) (0.014) (0.014) Constant 0.389*** 0.401*** 0.369*** 0.379*** 0.367*** 0.349*** (0.050) (0.046) (0.051) (0.055) (0.060) (0.063) Observations 384 384 358 358 358 358 R-squared 0.252 0.384 0.452 0.453 0.452 0.462 Note: Robust standard errors in parentheses. * P < 0.1; ** P < 0.05; *** P < 0.01. Table 5. Linear probability model predicting integration probability with integration experience Variables (1) (2) (3) (4) (5) (6) Integration Integration Integration Integration Integration Integration Integration experience 0.138*** 0.164*** 0.163*** 0.166*** 0.188*** (0.018) (0.019) (0.019) (0.038) (0.037) Knowledge codification 0.027 0.042 0.028 −0.007 (0.020) (0.034) (0.026) (0.047) Integration experience × Knowledge codification −0.001 0.059 (0.020) (0.036) Knowledge codification ^ 2 −0.012 0.021 (0.021) (0.026) Integration experience × Knowledge codification ^ 2 −0.035** (0.017) Acquisition experience −0.002 −0.009*** −0.007*** −0.007*** −0.007*** −0.008*** (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) Relatedness 0.339*** 0.300*** 0.313*** 0.313*** 0.313*** 0.320*** (0.050) (0.047) (0.049) (0.049) (0.049) (0.049) Parallel acquisitions 0.034*** 0.020*** 0.002 0.003 0.002 0.005 (0.006) (0.006) (0.006) (0.006) (0.007) (0.007) Target quality −0.066*** −0.044** −0.060*** −0.058*** −0.059*** −0.060*** (0.017) (0.018) (0.019) (0.018) (0.019) (0.019) Relative acquisition size 0.012 0.006 −0.008 −0.008 −0.008 −0.008 (0.012) (0.010) (0.014) (0.014) (0.014) (0.014) Constant 0.389*** 0.401*** 0.369*** 0.379*** 0.367*** 0.349*** (0.050) (0.046) (0.051) (0.055) (0.060) (0.063) Observations 384 384 358 358 358 358 R-squared 0.252 0.384 0.452 0.453 0.452 0.462 Variables (1) (2) (3) (4) (5) (6) Integration Integration Integration Integration Integration Integration Integration experience 0.138*** 0.164*** 0.163*** 0.166*** 0.188*** (0.018) (0.019) (0.019) (0.038) (0.037) Knowledge codification 0.027 0.042 0.028 −0.007 (0.020) (0.034) (0.026) (0.047) Integration experience × Knowledge codification −0.001 0.059 (0.020) (0.036) Knowledge codification ^ 2 −0.012 0.021 (0.021) (0.026) Integration experience × Knowledge codification ^ 2 −0.035** (0.017) Acquisition experience −0.002 −0.009*** −0.007*** −0.007*** −0.007*** −0.008*** (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) Relatedness 0.339*** 0.300*** 0.313*** 0.313*** 0.313*** 0.320*** (0.050) (0.047) (0.049) (0.049) (0.049) (0.049) Parallel acquisitions 0.034*** 0.020*** 0.002 0.003 0.002 0.005 (0.006) (0.006) (0.006) (0.006) (0.007) (0.007) Target quality −0.066*** −0.044** −0.060*** −0.058*** −0.059*** −0.060*** (0.017) (0.018) (0.019) (0.018) (0.019) (0.019) Relative acquisition size 0.012 0.006 −0.008 −0.008 −0.008 −0.008 (0.012) (0.010) (0.014) (0.014) (0.014) (0.014) Constant 0.389*** 0.401*** 0.369*** 0.379*** 0.367*** 0.349*** (0.050) (0.046) (0.051) (0.055) (0.060) (0.063) Observations 384 384 358 358 358 358 R-squared 0.252 0.384 0.452 0.453 0.452 0.462 Note: Robust standard errors in parentheses. * P < 0.1; ** P < 0.05; *** P < 0.01. Table 6. Linear probability model predicting integration probability with the PoIAs Variables (1) (2) (3) (4) (5) (6) Integration Integration Integration Integration Integration Integration PoIAs 0.432*** 0.506*** 0.504*** 0.345*** 0.445*** (0.055) (0.059) (0.059) (0.099) (0.104) Knowledge codification 0.016 0.038 −0.044 −0.093 (0.022) (0.035) (0.035) (0.061) PoIA × knowledge codification 0.128** 0.277*** (0.059) (0.100) Knowledge codification ^ 2 −0.018 0.034 (0.021) (0.035) PoIA × knowledge codification ^ 2 −0.109** (0.055) Acquisition experience −0.001 0.000 0.003 0.004* 0.004* 0.004* (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) Relatedness 0.342*** 0.290*** 0.306*** 0.306*** 0.304*** 0.318*** (0.048) (0.047) (0.049) (0.049) (0.049) (0.049) Parallel acquisitions 0.035*** 0.022*** 0.006 0.007 −0.000 0.004 (0.006) (0.006) (0.006) (0.006) (0.007) (0.007) Target quality −0.067*** −0.060*** −0.076*** −0.074*** −0.075*** −0.075*** (0.017) (0.017) (0.018) (0.018) (0.018) (0.017) Relative acquisition size 0.015 0.003 −0.010 −0.010 −0.011 −0.010 (0.011) (0.010) (0.014) (0.014) (0.014) (0.014) Constant 0.378*** 0.239*** 0.186*** 0.202*** 0.267*** 0.223*** (0.048) (0.050) (0.054) (0.058) (0.068) (0.077) Observations 408 384 358 358 358 358 R-squared 0.254 0.373 0.432 0.433 0.442 0.450 Variables (1) (2) (3) (4) (5) (6) Integration Integration Integration Integration Integration Integration PoIAs 0.432*** 0.506*** 0.504*** 0.345*** 0.445*** (0.055) (0.059) (0.059) (0.099) (0.104) Knowledge codification 0.016 0.038 −0.044 −0.093 (0.022) (0.035) (0.035) (0.061) PoIA × knowledge codification 0.128** 0.277*** (0.059) (0.100) Knowledge codification ^ 2 −0.018 0.034 (0.021) (0.035) PoIA × knowledge codification ^ 2 −0.109** (0.055) Acquisition experience −0.001 0.000 0.003 0.004* 0.004* 0.004* (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) Relatedness 0.342*** 0.290*** 0.306*** 0.306*** 0.304*** 0.318*** (0.048) (0.047) (0.049) (0.049) (0.049) (0.049) Parallel acquisitions 0.035*** 0.022*** 0.006 0.007 −0.000 0.004 (0.006) (0.006) (0.006) (0.006) (0.007) (0.007) Target quality −0.067*** −0.060*** −0.076*** −0.074*** −0.075*** −0.075*** (0.017) (0.017) (0.018) (0.018) (0.018) (0.017) Relative acquisition size 0.015 0.003 −0.010 −0.010 −0.011 −0.010 (0.011) (0.010) (0.014) (0.014) (0.014) (0.014) Constant 0.378*** 0.239*** 0.186*** 0.202*** 0.267*** 0.223*** (0.048) (0.050) (0.054) (0.058) (0.068) (0.077) Observations 408 384 358 358 358 358 R-squared 0.254 0.373 0.432 0.433 0.442 0.450 Note: Robust standard errors in parentheses. * P < 0.1; ** P < 0.05; *** P < 0.01. Table 6. Linear probability model predicting integration probability with the PoIAs Variables (1) (2) (3) (4) (5) (6) Integration Integration Integration Integration Integration Integration PoIAs 0.432*** 0.506*** 0.504*** 0.345*** 0.445*** (0.055) (0.059) (0.059) (0.099) (0.104) Knowledge codification 0.016 0.038 −0.044 −0.093 (0.022) (0.035) (0.035) (0.061) PoIA × knowledge codification 0.128** 0.277*** (0.059) (0.100) Knowledge codification ^ 2 −0.018 0.034 (0.021) (0.035) PoIA × knowledge codification ^ 2 −0.109** (0.055) Acquisition experience −0.001 0.000 0.003 0.004* 0.004* 0.004* (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) Relatedness 0.342*** 0.290*** 0.306*** 0.306*** 0.304*** 0.318*** (0.048) (0.047) (0.049) (0.049) (0.049) (0.049) Parallel acquisitions 0.035*** 0.022*** 0.006 0.007 −0.000 0.004 (0.006) (0.006) (0.006) (0.006) (0.007) (0.007) Target quality −0.067*** −0.060*** −0.076*** −0.074*** −0.075*** −0.075*** (0.017) (0.017) (0.018) (0.018) (0.018) (0.017) Relative acquisition size 0.015 0.003 −0.010 −0.010 −0.011 −0.010 (0.011) (0.010) (0.014) (0.014) (0.014) (0.014) Constant 0.378*** 0.239*** 0.186*** 0.202*** 0.267*** 0.223*** (0.048) (0.050) (0.054) (0.058) (0.068) (0.077) Observations 408 384 358 358 358 358 R-squared 0.254 0.373 0.432 0.433 0.442 0.450 Variables (1) (2) (3) (4) (5) (6) Integration Integration Integration Integration Integration Integration PoIAs 0.432*** 0.506*** 0.504*** 0.345*** 0.445*** (0.055) (0.059) (0.059) (0.099) (0.104) Knowledge codification 0.016 0.038 −0.044 −0.093 (0.022) (0.035) (0.035) (0.061) PoIA × knowledge codification 0.128** 0.277*** (0.059) (0.100) Knowledge codification ^ 2 −0.018 0.034 (0.021) (0.035) PoIA × knowledge codification ^ 2 −0.109** (0.055) Acquisition experience −0.001 0.000 0.003 0.004* 0.004* 0.004* (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) Relatedness 0.342*** 0.290*** 0.306*** 0.306*** 0.304*** 0.318*** (0.048) (0.047) (0.049) (0.049) (0.049) (0.049) Parallel acquisitions 0.035*** 0.022*** 0.006 0.007 −0.000 0.004 (0.006) (0.006) (0.006) (0.006) (0.007) (0.007) Target quality −0.067*** −0.060*** −0.076*** −0.074*** −0.075*** −0.075*** (0.017) (0.017) (0.018) (0.018) (0.018) (0.017) Relative acquisition size 0.015 0.003 −0.010 −0.010 −0.011 −0.010 (0.011) (0.010) (0.014) (0.014) (0.014) (0.014) Constant 0.378*** 0.239*** 0.186*** 0.202*** 0.267*** 0.223*** (0.048) (0.050) (0.054) (0.058) (0.068) (0.077) Observations 408 384 358 358 358 358 R-squared 0.254 0.373 0.432 0.433 0.442 0.450 Note: Robust standard errors in parentheses. * P < 0.1; ** P < 0.05; *** P < 0.01. Figure 1. View largeDownload slide Moderating effect of knowledge codification on the relationship between integration experience and integration (based on Model 6 in Table 5). Figure 1. View largeDownload slide Moderating effect of knowledge codification on the relationship between integration experience and integration (based on Model 6 in Table 5). Finally, we study how persistence in acquisition integration is intertwined with momentum in acquisition. In particular, we study whether the persistence in decision-making generated by integration experience increases as acquisition experience increases. Given that both acquisition and integration experience are likely to be subject to routinization processes, they could reinforce each other as they increase more. Using a logistic regression, Model 1 in Table 7 shows that the interaction term between integration experience and acquisition experience is not significant, suggesting that the impact of integration experience on the probability of integrating a newly acquired target does not depend on acquisition experience. We find similar results if we operationalize integration experience using the PoIA. Model 2 in Table 7 shows that the interaction term between the PoIA and integration experience is not significant, which suggests that the inertia caused by integration experience does not depend on the level of acquisition experience. Table 7. Logit predicting integration probability with the interaction between integration experience and acquisition experience (Model 1) or the PoIA and acquisition experience (Model 2) (1) (2) Variables Integration Integration Integration experience 0.923*** (0.223) Acquisition experience −0.049* 0.004 (0.029) (0.025) Integration experience × acquisition experience 0.002 (0.015) PoIA 3.178*** (0.768) PoIA × acquisition experience 0.033 (0.061) Relatedness 2.135*** 2.090*** (0.358) (0.354) Parallel acquisitions 0.369*** 0.398*** (0.083) (0.085) Target quality −0.356* −0.487*** (0.185) (0.180) Relative acquisition size 0.020 −0.023 (0.072) (0.075) Constant −1.555*** −2.799*** (0.423) (0.545) Observations 384 384 Pseudo R-squared 0.3980 0.4081 Log likelihood −130.647 −128.460 (1) (2) Variables Integration Integration Integration experience 0.923*** (0.223) Acquisition experience −0.049* 0.004 (0.029) (0.025) Integration experience × acquisition experience 0.002 (0.015) PoIA 3.178*** (0.768) PoIA × acquisition experience 0.033 (0.061) Relatedness 2.135*** 2.090*** (0.358) (0.354) Parallel acquisitions 0.369*** 0.398*** (0.083) (0.085) Target quality −0.356* −0.487*** (0.185) (0.180) Relative acquisition size 0.020 −0.023 (0.072) (0.075) Constant −1.555*** −2.799*** (0.423) (0.545) Observations 384 384 Pseudo R-squared 0.3980 0.4081 Log likelihood −130.647 −128.460 Note: Robust standard errors in parentheses. * P < 0.1; ** P < 0.05; *** P < 0.01. Table 7. Logit predicting integration probability with the interaction between integration experience and acquisition experience (Model 1) or the PoIA and acquisition experience (Model 2) (1) (2) Variables Integration Integration Integration experience 0.923*** (0.223) Acquisition experience −0.049* 0.004 (0.029) (0.025) Integration experience × acquisition experience 0.002 (0.015) PoIA 3.178*** (0.768) PoIA × acquisition experience 0.033 (0.061) Relatedness 2.135*** 2.090*** (0.358) (0.354) Parallel acquisitions 0.369*** 0.398*** (0.083) (0.085) Target quality −0.356* −0.487*** (0.185) (0.180) Relative acquisition size 0.020 −0.023 (0.072) (0.075) Constant −1.555*** −2.799*** (0.423) (0.545) Observations 384 384 Pseudo R-squared 0.3980 0.4081 Log likelihood −130.647 −128.460 (1) (2) Variables Integration Integration Integration experience 0.923*** (0.223) Acquisition experience −0.049* 0.004 (0.029) (0.025) Integration experience × acquisition experience 0.002 (0.015) PoIA 3.178*** (0.768) PoIA × acquisition experience 0.033 (0.061) Relatedness 2.135*** 2.090*** (0.358) (0.354) Parallel acquisitions 0.369*** 0.398*** (0.083) (0.085) Target quality −0.356* −0.487*** (0.185) (0.180) Relative acquisition size 0.020 −0.023 (0.072) (0.075) Constant −1.555*** −2.799*** (0.423) (0.545) Observations 384 384 Pseudo R-squared 0.3980 0.4081 Log likelihood −130.647 −128.460 Note: Robust standard errors in parentheses. * P < 0.1; ** P < 0.05; *** P < 0.01. 5. Discussion In this article, we have developed and tested a theory of how past decisional persistence and knowledge codification affect persistence in decision-making. We considered two fundamental factors related to learning—integration experience and knowledge codification—and we theorized and showed that whereas integration experience tends to generate persistence in decision-making, codification of knowledge of the integration process amplifies this persistence at low levels and then overcomes it at high levels. Moreover, we provided evidence that implies that persistence in decision-making is likely not related to learning—i.e., superior capabilities—but rather to inertia—i.e., persistence with a decision despite its negative performance consequences. This study offers several contributions. First, we help explain why a growing body of research on the performance implications of experience in strategic contexts offers mixed results (Barkema and Schijven, 2008). By analyzing how experience and knowledge codification interact, we avoid the common tendency to infer about the impact of experience simply by looking at its direct impact on the nature of subsequent decisions or performance. Instead, we propose and show that the impact of experience on decision-making persistence is contingent on other factors, such as knowledge codification; even more importantly, this moderating effect is nonlinear. This finding suggests that future research should look further at the contingency factors of experience to shed new light on the mixed results of experience on subsequent decisions and performance. Second, through our investigation, we respond to calls for a stronger focus on organizational theory within decision-making research. Gavetti and Rivkin (2007) argue that “the centrality of both organizations and decision making has been, to a great extent, lost” and that “work in organizational learning, despite having clear roots in the Carnegie School, has shifted away from issues of decision making” (p. 524). By studying the impact of experience and knowledge codification on the persistence of decisions, we retrain the spotlight on the relationship between decision-making and two fundamental factors of organizational learning. Third, in analyzing the effects of knowledge codification at different levels, we show that knowledge codification cannot be characterized solely as a source of inertia (i.e., coercive factor) or as a source of learning (i.e., enabling factor). Instead, knowledge codification likely changes from a magnifier of inertia to a source of learning as it increases. These results thus challenge the conventional wisdom that “portray[s] codification as a panacea, largely disregarding the inertia it entails, as stressed by a long line of prior work” (Heimeriks et al., 2012: 720). Rather, we find that codification simultaneously has pros and cons and that its inertial effects are overcome after a certain level. At the extreme, organizations without any knowledge are thus less likely to fall into decision inertia traps than those with a little knowledge codification. However, our findings also show that significant investments in knowledge overcome these inertial effects. Finally, our article is also germane to the literature that frames the effect of organizational acquisition experience on acquisition performance from a behavioral learning perspective (Haleblian and Finkelstein, 1999; Finkelstein and Haleblian, 2002). These studies have documented a U-shaped relationship between the number of acquisitions a firm has conducted and a focal acquisition’s performance. The theory behind this result hinges on a “negative transfer” process. After making their first acquisitions, inexperienced acquirers inappropriately generalize what they have learned to subsequent acquisitions, which are different from those previously conducted, thus hurting acquisition performance. More-experienced acquirers, however, learn to appropriately discriminate between their acquisitions and are less likely to unduly transfer routines and practices established in prior situations to new, dissimilar situations. Although this theory is appealing and has received empirical confirmation, it has investigated mostly the (indirect) link between experience and performance. In this study, we attempt to dig deeper into this process and study the effects of experiential and deliberate learning on the organizational decision process and how they affect performance in turn. We obtain results consistent with this theory. Thus, our results provide additional evidence consistent with the idea that firms tend to persist in using and applying old schemes and routines when making decisions related to organizational integration in the post-M&A phase. Nonetheless, complementing prior studies, we show that knowledge codification is the actual mechanism behind firms’ ability to recognize two different deals as different, and hence, it helps firms avoid the “negative transfer” effect caused by the accumulation of experience with a particular decision. Our results also have important managerial implications. First, acquirers that invest in knowledge codification can learn to overcome persistence in post-acquisition decisions. Therefore, managers should realize that increasing their firms’ knowledge levels decreases automatic triggers of predefined responses, reduces inertia, and encourages more-varied, effective decision-making processes. In addition, in the post-acquisition phase, managerial decisions need to be tailored to the focal acquisition. Moreover, the finding that the benefits of knowledge codification outweigh its costs only when a firm has sufficiently invested in knowledge codification is particularly important for acquiring firms. Acquirers vary significantly in the extent to which they develop specific tools based on their post-acquisition integration activities: some acquirers codify most activities in manuals; others codify few of them. An acquiring firm that has developed only a due diligence checklist, for example, differs greatly from one that has developed multiple manuals to describe the different activities in the post-merger phase, such as the conversion of information systems, the affiliation of human resources, financial evaluation spreadsheets, product training programs, and project management packages. Our findings suggest that investing in knowledge codification is worthwhile only if firms span a significant number of post-acquisition integration activities. However, such extensive codification is challenging for many acquirers, such as small firms with few resources to invest in developing manuals or those that acquire only a few targets throughout their history. Therefore, it would be interesting to investigate whether the difficulty they face in developing complete codifications of different M&A activities might be overcome, at least partially, by involving investment bankers, consulting firms, or strategic advisors. Our findings, however, suggest that it is not beneficial to begin a codification process superficially and merely following managerial fads. Our study has a number of limitations, some of which refer to the validity and generalizability of our results. In this single-industry study, we focus only on US bank mergers. As usual, the applicability of the findings to other samples in the same industry, other industries, or other geographic and institutional contexts should therefore be considered carefully. Other limitations arise from the need to triangulate and validate our results across different post-acquisition decisions. Additionally, acquisitions represent a particularly rare and complex type of strategic decision. Future studies should analyze whether similar results arise for strategic decisions that are less rare or complex. Most importantly, our data structure does not permit us to unambiguously establish causality, only conditional correlations. Finally, because of a lack of data, we do not study the role of knowledge articulation (Zollo and Winter, 2002). Studying whether knowledge articulation strengthens or weakens the persistence caused by integration experience represents an important avenue for future research. Nevertheless, our results can guide scholars in promising directions, namely, toward a better understanding of the antecedents and evolution of organizational decisions. In addition, we hope further research can provide new insights on the causes of persistence in decision-making and the learning mechanisms that might mitigate these forces. Researchers might also shed new light on the micro-processes that explain why knowledge codification generates inertia below a certain threshold and learning above it. Acknowledgments We are grateful to editor Fredrik Tell and two anonymous reviewers for their insightful comments and guidance. We also would like to thank Stefano Brusoni, Raffaele Conti, Phanish Puranam, as well as seminar participants at the SEI Faculty Workshop and the AOM Meeting for valuable comments on previous versions of the manuscript. Francesco Castellaneta acknowledges funding from Catolica Lisbon, and Giovanni Valentini from the “Agencia Estatal de Investigación” (AEI) of the Spanish Ministry of Economy and Competitiveness (Ref. ECO2015-ECO2015-71173-P) (AEI/FEDER, UE). Footnotes 1 The data set we use in this article has been used in two other published papers: (Zollo, 2009) and (Zollo and Singh, 2004). However, the research questions pursued in these two papers clearly differ from the one addressed in this article. The two papers explain M&A performance (i.e., M&A performance is the dependent variable), while in this article, we explain the post-acquisition integration decision (i.e., the integration decision is the dependent variable) instead. 2 In robustness results (available upon request), we do not dichotomize the variable but use a dependent variable that takes the following values: 0 if “few or no features were aligned or centralized”; 1 “if only selected systems, procedures, or products were aligned or centralized”; 2 if “many but not all systems, procedures, and products were aligned or centralized”; and 3 if “all systems, procedures, and products were completely integrated.” When using an ordered probit model, we obtain similar results. 3 To avoid zero values for the argument of the log, we follow Helsel (2005) and add the smallest quantity that makes the lognormal probability plot approximately linear. 4 Results (available upon request) are robust to the inclusion of year dummies. 5 Due to missing values for ROA, we include in the analysis only 26 acquiring banks that completed 259 acquisitions overall. 6 This result is indirectly clear from previous regressions but confirmed when directly regressing Deviation on integration experience (results available from the authors). References Adler P. S. , Borys B. ( 1996 ), ‘ Two types of bureaucracy: enabling and coercive ,’ Administrative Science Quarterly , 41 ( 1 ), 61 – 89 . Google Scholar CrossRef Search ADS Amburgey T. , Miner A. ( 1992 ), ‘ Strategic momentum: the effects of repetitive, positional, and contextual momentum on merger activity ,’ Strategic Management Journal , 13 ( 5 ), 335 – 348 . Google Scholar CrossRef Search ADS Amburgey T. L. , Kelly D. , Barnett W. P. ( 1993 ), ‘ Resetting the clock: the dynamics of organizational change and failure ,’ Administrative Science Quarterly , 38 ( 1 ), 51 – 73 . Google Scholar CrossRef Search ADS Balconi M. , Pozzali A. , Viale R. ( 2007 ), ‘ The ‘codification debate’ revisited: a conceptual framework to analyze the role of tacit knowledge in economics ,’ Industrial and Corporate Change , 16 ( 5 ), 823 – 849 . Google Scholar CrossRef Search ADS Barkema H. G. , Schijven M. ( 2008 ), ‘ How do firms learn to make acquisitions? A review of past research and an agenda for the future ,’ Journal of Management , 34 ( 3 ), 594 – 634 . Google Scholar CrossRef Search ADS Baum J. A. , Li S. X. , Usher J. M. ( 2000 ), ‘ Making the next move: how experiential and vicarious learning shape the locations of chains’ acquisitions ,’ Administrative Science Quarterly , 45 ( 4 ), 766 – 801 . Google Scholar CrossRef Search ADS Becker M. C. ( 2004 ), ‘ Organizational routines: a review of the literature ,’ Industrial and Corporate Change , 13 ( 4 ), 643 – 678 . Google Scholar CrossRef Search ADS Becker M. C. , Lazaric N. , Nelson R. , Winter S. ( 2005 ), ‘ Applying organizational routines in understanding organizational change ,’ Industrial and Corporate Change , 14 ( 5 ), 775 – 791 . Google Scholar CrossRef Search ADS Benner M. J. , Tushman M. ( 2002 ), ‘ Process management and technological innovation: a longitudinal study of the photography and paint industries ,’ Administrative Science Quarterly , 47 ( 4 ), 676 – 707 . Google Scholar CrossRef Search ADS Bergh D. D. , Lim E. N. ( 2008 ), ‘ Learning how to restructure: absorptive capacity and improvisational views of restructuring actions and performance ,’ Strategic Management Journal , 29 ( 6 ), 593 – 616 . Google Scholar CrossRef Search ADS Castellaneta F. , Salvato C. ( forthcoming ), ‘ Culminating events and time working together in top management teams: insights from private equity ,’ Long Range Planning . Castellaneta F. , Zollo M. ( 2015 ), ‘ The dimensions of experiential learning in the management of activity load ,’ Organization Science , 26 ( 1 ), 140 – 157 . Google Scholar CrossRef Search ADS Cowan R. , David P. A. , Foray D. ( 2000 ), ‘ The explicit economics of knowledge codification and tacitness ,’ Industrial and Corporate Change , 9 ( 2 ), 211 – 253 . Google Scholar CrossRef Search ADS Cowan R. , Foray D. ( 1997 ), ‘ The economics of codification and the diffusion of knowledge ,’ Industrial and Corporate Change , 6 ( 3 ), 595 – 622 . Google Scholar CrossRef Search ADS Cyert R. M. , March J. G. ( 1963 ), A Behavioral Theory of the Firm . Prentice Hall : New York, NY . Datta D. K. ( 1991 ), ‘ Organizational fit and acquisition performance: effects of post-acquisition integration ,’ Strategic Management Journal , 12 ( 4 ), 281 – 297 . Google Scholar CrossRef Search ADS DeYoung R. , Evanoff D. D. , Molyneux P. ( 2009 ), ‘ Mergers and acquisitions of financial institutions: a review of the post-2000 literature ,’ Journal of Financial Services Research , 36 ( 2–3 ), 87 – 110 . Google Scholar CrossRef Search ADS Dimov D. , de Holan P. M. , Milanov H. ( 2012 ), ‘ Learning patterns in venture capital investing in new industries ,’ Industrial and Corporate Change , 21 ( 6 ), 1389 – 1426 . Google Scholar CrossRef Search ADS Finkelstein S. , Haleblian J. ( 2002 ), ‘ Understanding acquisition performance: the role of transfer effects ,’ Organization Science , 13 ( 1 ), 36 – 47 . Google Scholar CrossRef Search ADS Foray D. , Steinmueller W. E. ( 2003 ), ‘ The economics of knowledge reproduction by inscription ,’ Industrial and Corporate Change , 12 ( 2 ), 299 – 319 . Google Scholar CrossRef Search ADS Gavetti G. , Levinthal D. A. ( 2000 ), ‘ Looking forward and looking backward: cognitive and experiential search ,’ Administrative Science Quarterly , 45 ( 1 ), 113 – 137 . Google Scholar CrossRef Search ADS Gavetti G. , Rivkin J. W. ( 2007 ), ‘ On the origin of strategy: action and cognition over time ,’ Organization Science , 18 ( 3 ), 420 – 439 . Google Scholar CrossRef Search ADS Groves R. M. , Cialdini R. B. , Couper M. P. ( 1992 ), ‘ Understanding the decision to participate in a survey ,’ Public Opinion Quarterly , 56 ( 4 ), 475 – 495 . Google Scholar CrossRef Search ADS Gulati R. ( 1995 ), ‘ Social structure and alliance formation patterns: a longitudinal analysis ,’ Administrative Science Quarterly , 40 ( 4 ), 619 – 652 . Google Scholar CrossRef Search ADS Gulati R. , Gargiulo M. ( 1999 ), ‘ Where do interorganizational networks come from? ,’ American Journal of Sociology , 104 ( 5 ), 1439 – 1493 . Google Scholar CrossRef Search ADS Haans R. F. , Pieters C. , He Z. ( 2016 ), ‘ Thinking about U: theorizing and testing U and inverted U‐shaped relationships in strategy research ,’ Strategic Management Journal , 37 ( 7 ), 1177 – 1195 . Google Scholar CrossRef Search ADS Hagendorff J. , Keasey K. ( 2009 ), ‘ Post-merger strategy and performance: evidence from the US and European banking industries ,’ Accounting and Finance , 49 ( 4 ), 725 – 751 . Google Scholar CrossRef Search ADS Håkanson L. ( 2007 ), ‘ Creating knowledge: the power and logic of articulation ,’ Industrial and Corporate Change , 16 ( 1 ), 51 – 88 . Google Scholar CrossRef Search ADS Haleblian J. , Finkelstein S. ( 1999 ), ‘ The influence of organizational acquisition experience on acquisition performance: a behavioral learning perspective ,’ Administrative Science Quarterly , 44 ( 1 ), 29 – 56 . Google Scholar CrossRef Search ADS Haleblian J. , Kim J. J. , Rajagopalan N. ( 2006 ), ‘ The influence of acquisition experience and performance on acquisition behavior: evidence from the U.S. Commercial banking industry ,’ Academy of Management Journal , 49 ( 2 ), 357 – 370 . Google Scholar CrossRef Search ADS Hannan T. H. , Pilloff S. J. ( 2009 ), ‘ Acquisition targets and motives in the banking industry ,’ Journal of Money, Credit and Banking , 41 ( 6 ), 1167 – 1187 . Google Scholar CrossRef Search ADS Heimeriks K. H. , Schijven M. , Gates S. ( 2012 ), ‘ Manifestations of higher-order routines: the underlying mechanisms of deliberate learning in the context of postacquisition integration ,’ Academy of Management Journal , 55 ( 3 ), 703 – 726 . Google Scholar CrossRef Search ADS Helsel D. R. ( 2005 ), Nondetects and Data Analysis . John Wiley and Sons : New York, NY . Horrace W. C. , Oaxaca R. L. ( 2006 ), ‘ Results on the bias and inconsistency of ordinary least squares for the linear probability model ,’ Economics Letters , 90 ( 3 ), 321 – 327 . Google Scholar CrossRef Search ADS Hoskisson R. E. , Hitt M. A. ( 1988 ), ‘ Strategic control systems and relative R&D investment in large multiproduct firms ,’ Strategic Management Journal , 9 ( 1 ), 605 – 621 . Google Scholar CrossRef Search ADS Kale P. , Singh H. ( 2007 ), ‘ Building firm capabilities through learning: the role of the alliance learning process in alliance capability and firm-level alliance success ,’ Strategic Management Journal , 28 ( 10 ), 981 – 1000 . Google Scholar CrossRef Search ADS Katz M. , Shapiro C. ( 1985 ), ‘ Network externalities, competition and compatibility ,’ American Economic Review , 75 ( 3 ), 424 – 440 . Levinthal D. A. , March J. G. ( 1993 ), ‘ The myopia of learning ,’ Strategic Management Journal , 14 ( S2 ), 95 – 112 . Google Scholar CrossRef Search ADS Levitt B. , March J. G. ( 1988 ), ‘ Organizational learning ,’ Annual Review of Sociology , 14 ( 1 ), 319 – 340 . Google Scholar CrossRef Search ADS March J. , Simon H. A. ( 1958 ), Organizations . Wiley : New York, NY . Nelson R. R. , Winter S. G. ( 1982 ), An Evolutionary Theory of Economic Change . Belknap : Cambridge, MA . Nonaka I. ( 1994 ), ‘ A dynamic theory of organizational knowledge creation ,’ Organization Science , 5 ( 1 ), 14 – 37 . Google Scholar CrossRef Search ADS Pentland B. T. , Feldman M. S. ( 2005 ), ‘ Organizational routines as a unit of analysis ,’ Industrial and Corporate Change , 14 ( 5 ), 793 – 815 . Google Scholar CrossRef Search ADS Prencipe A. , Tell F. ( 2001 ), ‘ Inter-project learning: processes and outcomes of knowledge codification in project-based firms ,’ Research Policy , 30 ( 9 ), 1373 – 1394 . Google Scholar CrossRef Search ADS Romme G. L. , Zollo M. , Berends P. ( 2010 ), ‘ Dynamic capabilities, deliberate learning and environmental dynamism: a simulation model ,’ Industrial and Corporate Change , 19 ( 4 ), 1271 – 1299 . Google Scholar CrossRef Search ADS Schulz M. ( 1998 ), ‘ Limits to bureaucratic growth: the density dependence of organizational rule births ,’ Administrative Science Quarterly , 43 ( 4 ), 845 – 876 . Google Scholar CrossRef Search ADS Shaver J. M. , Mitchell W. , Yeung B. ( 1997 ), ‘ The effect of own-firm and other-firm experience on foreign direct investment survival in the United States, 1987–1992 ,’ Strategic Management Journal , 18 ( 10 ), 811 – 824 . Google Scholar CrossRef Search ADS Snell S. A. ( 1992 ), ‘ Control theory in strategic human resource management: the mediating effect of administrative information ,’ Academy of Management Journal , 35 ( 2 ), 292 – 327 . Sydow J. , Schreyögg G. , Koch J. ( 2009 ), ‘ Organizational path dependence: opening the black box ,’ Academy of Management Review , 34 ( 4 ), 689 – 709 . Szulanski G. ( 1996 ), ‘ Exploring internal stickiness: impediments to the transfer of best practice within the firm ,’ Strategic Management Journal , 17 ( 1 ), 27 – 43 . Google Scholar CrossRef Search ADS Zander U. , Kogut B. ( 1995 ), ‘ Knowledge and the speed of the transfer and imitation of organizational capabilities: an empirical test ,’ Organization Science , 6 ( 1 ), 76 – 92 . Google Scholar CrossRef Search ADS Zollo M. ( 2009 ), ‘ Superstitious learning with rare strategic decisions: theory and evidence from corporate acquisitions ,’ Organization Science , 20 ( 5 ), 894 – 908 . Google Scholar CrossRef Search ADS Zollo M. , Singh H. ( 2004 ), ‘ Deliberate learning in corporate acquisitions: post-acquisition strategies and integration capability in U.S. bank mergers ,’ Strategic Management Journal , 25 ( 13 ), 1233 – 1256 . Google Scholar CrossRef Search ADS Zollo M. , Winter S. G. ( 2002 ), ‘ Deliberate learning and the evolution of dynamic capabilities ,’ Organization Science , 13 ( 3 ), 339 – 351 . Google Scholar CrossRef Search ADS © The Author 2017. Published by Oxford University Press on behalf of Associazione ICC. All rights reserved. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Industrial and Corporate Change Oxford University Press

Learning or inertia? The impact of experience and knowledge codification on post-acquisition integration

Loading next page...
 
/lp/ou_press/learning-or-inertia-the-impact-of-experience-and-knowledge-J6cog4I60w
Publisher
Oxford University Press
Copyright
© The Author 2017. Published by Oxford University Press on behalf of Associazione ICC. All rights reserved.
ISSN
0960-6491
eISSN
1464-3650
D.O.I.
10.1093/icc/dtx043
Publisher site
See Article on Publisher Site

Abstract

Abstract This article develops and tests a theory on the evolution of complex organizational decisions, such as the decision to integrate (or not) a target company during the post-acquisition management phase. Using a sample of US bank mergers, we show that persistence in—or variation of—integration decisions depends on two key factors: integration experience and related knowledge codification. Integration experience tends to generate persistence in the integration decision, which is associated with poor deal performance. Moreover, we show that when knowledge codification is low, the persistence caused by integration experience further increases. However, when knowledge codification is sufficiently high, the inertial effect of experience diminishes significantly. Hence, high levels of knowledge codification can weaken the effects of decision inertia, which suggests that as knowledge codification increases, the role of knowledge codification switches from strengthening inertia to promoting learning. 1. Introduction Ample empirical evidence suggests that as an organization gains experience with a particular decision, it becomes more likely to repeat it (Amburgey et al., 1993; Gulati, 1995; Shaver et al., 1997; Dimov et al., 2012). Underlying this perspective is the assumption that experience facilitates the formation and refinement of routines (Levitt and March, 1988). Routines are programs of action that, reflecting an organization’s prior experience with a particular decision or task (Nelson and Winter, 1982), may constrain the organization’s future behavior (March and Simon, 1958; Cyert and March, 1963) and generate inertial decision-making (Szulanski, 1996). This may also happen in the context of strategic decisions, where routines often “play a crucial rule in the formulation of a firm’s strategic choices by supplementing, or even substituting for, calculative, formal strategic decision-making rules” (Haleblian et al., 2006: 358). In this article, we first closely follow the prior literature and establish our baseline hypothesis, arguing that as organizational experience with a given decision increases, the likelihood of repeating that decision in the future increases. When persistence in decision-making is caused by inertia, the stability-proving effect of routines can develop into a pathology (Becker, 2004). Based on the understanding that persistence in decision-making can be harmful, we therefore study how knowledge codification—the codification of related knowledge in written tools such as manuals, blueprints, spreadsheets, and decision support systems (Zollo and Winter, 2002)—may affect this persistence. Note that whereas knowledge codification and experience are likely to covary, they represent two different concepts because “not all codifiable … knowledge is actually codified” (Zollo and Singh, 2004: 1238). Therefore, firms with equal levels of experience might have different levels of knowledge codification (Romme et al., 2010). There are various perspectives on the possible effects of codification. A stream of literature contends that knowledge codification results in true organizational learning, which should remedy the inertial persistence in decision-making. In particular, knowledge codification, which requires active and intentional sense-making processes and a significant cognitive investment, generates a deep understanding of past experience and therefore diminishes the likelihood of inappropriate generalizations (Finkelstein and Haleblian, 2002; Zollo and Winter, 2002). In turn, knowledge codification should bring greater organizational adaptation and therefore reduce persistence in decision-making. Another stream of research has noted that the tools produced through knowledge codification may also act as a source of behavioral control that specifies how decision makers should act in a given situation based on prior experience (Hoskisson and Hitt, 1988; Snell, 1992). Therefore, the codification of knowledge in written tools may also reinforce inertial behavior. We reconcile these perspectives and argue that at low levels of codification, the “behavioral control” effect dominates, whereas when codification rises above a certain threshold, deeper learning of the cause-and-effect relationships takes place. We test these theoretical arguments by focusing on a strategic event common to most large organizations—mergers and acquisitions (M&As)—and more specifically on the decision of whether to integrate the target firm after acquiring it. In particular, we study how experience and knowledge codification regarding the integration decision affect the probability of integrating a newly acquired target firm. To this end, we use a sample of US bank mergers realized between 1985 and 1995. Although these data are relatively old, they are still appropriate for testing our theory for three reasons. First, the US commercial banking industry engaged in several acquisitions after 1985, and therefore, the conditions were ripe for the accumulation of experience and the codification of knowledge. Second, since 1996, US banks have pursued strategies (Hagendorff and Keasey, 2009; Hannan and Pilloff, 2009) that are likely to more consistently result in target integration, reducing the likelihood of observing variation in post-acquisition integration decisions. Third, the sample contains reliable (and typically rare) data on integration decisions and knowledge codification. Nevertheless, we believe that the process studied here is not specific to a given business cycle or industry and that our data, though not recent, are subject to the usual generalizability caveats of any empirical paper. Consistent with previous literature, our findings show that experience with the integration decision increases the likelihood of integrating the target firm in the following acquisition. Moreover, we find evidence suggesting that persistence in decision-making is less likely to be due to superior capabilities in the integration process than to inertia, which is consistent with the view that “the stability-providing effect of routines does […] develop into a pathology” (Becker, 2004: 659). Most importantly, we show that whereas knowledge codification initially increases persistence in decision-making, the effect reverses at high levels of codification, and codification diminishes the effect of prior experience, thus helping organizations to overcome inertia. These results offer two main contributions to the literature. First, on a general level, we advance the organizational learning literature by shedding new light on several contradictory claims and unclear results on the impact of knowledge codification, which has been characterized as a source of either inertia (i.e., a coercive factor) or learning (i.e., an enabling factor) (Adler and Borys, 1996). We reveal that the role of knowledge codification changes from coercive to enabling as knowledge codification increases. Therefore, knowledge codification could be a remedy to organizational inertia, but only after codification rises above a certain threshold. Second, on a more applied level, we advance the M&A literature by providing new evidence on the determinants of integration decisions that—although crucial for explaining acquisition success—have received relatively scant attention in prior research. 2. Theory and hypotheses 2.1 Integration experience and persistence in decision-making A number of empirical studies have shown that persistence in decision-making pervades organizational life, characterizing various decisions related to a firm’s corporate strategy. For example, Haleblian et al. (2006) find that as an acquirer’s acquisition experience increases, its likelihood to make a subsequent acquisition increases. Bergh and Lim (2008) document that as firms gain experience with sell-offs, they continue to use sell-offs as a preferred corporate restructuring strategy. Additionally, Gulati and Gargiulo (1999) show that previous ties between organizations increase the probability of a future alliance between them, while Baum et al. (2000) find that Canadian nursing home chains tend to acquire targets similar to their prior acquisitions. Overall, several findings in this stream of research provide consistent evidence that prior experience is an important predictor of future behaviors and that experience leads to persistence in strategic decision-making. The core insight is that this process occurs because of the emergence of routines (Nelson and Winter, 1982), which are of fundamental importance to understanding organizational change (Becker et al., 2005). Gavetti and Levinthal (2000) note that “routines reflect experiential wisdom in that they are the outcome of trial and error learning and the selection and retention of prior behaviors” (p. 113). Similarly, Pentland and Feldman (2005) contend that because “any particular organizational routine can exhibit a great deal of continuity over time” (p. 794), routines are often considered a source of organizational stability (see Becker, 2004 for a review). Persistence in decision-making might also affect the empirical context of this study, that is, post-acquisition integration decisions. Gavetti and Rivkin (2007) provide an interesting example in this respect. In 1997, the Internet portal Lycos completed its first acquisition, Tripod—a homepage site giving users software tools and server space to build their own Web pages. After much debate on how to handle Tripod’s acquisition, “the management team decided to keep the Tripod brand and integrate Tripod selectively … Lycos repeated this decision to maintain multiple brands during the subsequent years as it acquired other sites such as Guestworld, WhoWhere?, Internet Music Distribution, etc. in rapid succession” (Gavetti and Rivkin, 2007: 426). While research on organizational learning offers ample evidence that as an organization becomes more experienced with a particular decision it becomes more likely to repeat it, it remains unclear whether this persistence is caused by learning or by inertia. Persistence in decision-making can contribute to the formation of superior organizational capabilities and could therefore positively impact performance. For example, take a serial acquirer that has made several acquisitions in its history and has integrated the vast majority of the acquired targets. Through accumulated integration experience, this serial acquirer could develop integration routines that help it handle the complex post-merger integration process. Routines’ stability-providing effect is important for learning because this “stability provides a base against which to assess change, compare and learn” (Becker, 2004: 659). Moreover, stability in organizational routines leads to predictability, which in turn aids coordination (Cyert and March, 1963; Nelson and Winter, 1982). Therefore, as noted by Haleblian et al. (2006: 358), “routines can become a source of competitive advantage” and generate persistence in decision-making based on learning and superior capabilities. However, while routines are likely to generate learning in the context of simple and operational organizational tasks, their role in complex and strategic decisions is less clear and can actually become harmful (Zollo, 2009). As noted by Becker (2004: 659), “At times, the stability-providing effect of routines does, however, develop into a pathology.” Routines can contribute to the formation of fixed and inflexible organizational behaviors, leading an organization to develop a lock-in and inertial decision process (Sydow et al., 2009). Whereas persistence in decision-making should not be conceived as a state of total organizational rigidity, it de facto reduces the variance in possible behaviors and decisions, which can be particularly problematic when the decision at hand is complex, such as the post-acquisition integration decision. Based on this understanding and consistent with prior related literature, we predict that persistence in decision-making could be due to inertia and propose the following hypothesis: H1. Due to inertia, as an acquirer’s experience in integration increases, its probability of integrating a newly acquired target increases. 2.2 Integration experience and knowledge codification Although persistence in decision-making pervades organizational life and, as we argue, is likely due to inertia rather than to learning, to date, the literature provides only a limited understanding of the factors that may reduce it. Thus, in this study, we examine how knowledge codification affects the persistence in decision-making caused by the accumulation of integration experience. Knowledge codification refers to the codification of organizational “understandings of the performance implications of internal routines in written tools, such as manuals, blueprints, spreadsheets, decision support systems, project management software, etc.” (Zollo and Winter, 2002: 342). Once knowledge is codified, less tacit knowledge remains “idiosyncratic to a person or few people, and more of it is transformed into some systematic form that can be communicated” (Cowan and Foray, 1997: 595) over time and space at a low marginal cost (Foray and Steinmueller, 2003; Prencipe and Tell, 2001). Moreover, different types of knowledge—from the less complex to the more complex—can be contained in these tools: (i) know-what, or the content, the information, and the facts; (ii) know-how, or the methodology and the procedure; and (iii) know-why, or the rationale, principles, theories, and causalities (Foray and Steinmueller, 2003; Håkanson, 2007; Kale and Singh, 2007). Knowledge codification can influence decision-making persistence because of its two outputs. The first and simplest output of knowledge codification is the provision of ready-to-use solutions. Codified knowledge “is inscripted in a memorization medium, usually a document” (Balconi et al., 2007: 833) that “serves inter alia as a storage depository, as a reference point” (Cowan et al., 2000: 169). These codified tools make it easier for organizations to address recurring problems that present themselves over time with similar characteristics, and this effect is readily observable. The second output of knowledge codification is realized over time. The process of creating and updating codification tools implies an effort to truly understand the output of past actions, even when learning is not the deliberate goal (Zollo and Winter, 2002). Therefore, knowledge codification goes well beyond the output of ready-to-use solutions and can provide a more profound understanding of the criteria to consider when making complex decisions. In other words, decision makers—by involving themselves in codification effort—emerge with a crisper understanding of the causal linkages between decisions and outcomes, that is, what works and what fails, under what conditions, and—most importantly—why (Kale and Singh, 2007). The reason is that decision makers, by repeatedly undergoing codification efforts, can identify the strengths and weaknesses of existing codified tools and change them accordingly (Gavetti and Levinthal, 2000). Moreover, knowledge codification presupposes knowledge articulation (Håkanson, 2007), or “the process through which implicit knowledge is articulated through collective discussions, debriefing sessions, and performance evaluation processes” (Zollo and Winter, 2002: 341). Therefore, the process of knowledge codification produces learning that goes above and beyond the tools in which knowledge is codified. This effect, however, develops over time, and only after sufficient effort. These two outputs of knowledge codification can exert different, contrasting moderating effects on the positive relationship between experience and decision-making persistence (i.e., H1). On one hand, the first output—the production of ready-to-use solutions captured in documents—may reinforce the inertial effect of experience. Hoskisson and Hitt (1988) and Snell (1992) contend that codified tools represent a form of behavioral control that dictates how decision makers should act in a given situation. Codified tools are repositories of organizational memory (Cyert and March, 1963; Nonaka, 1994; Zander and Kogut, 1995) that are designed to ensure that everyone complies with established organizational practices. In this respect, the knowledge recorded in these tools can serve not only as a reference point in the decision-making process but also “possibly as an authority” (Cowan et al., 2000), even in cases different from those that led to codification. It has been argued that codified tools can do much more than provide a simple possible ready-to-use solution to problems the firm encounters; they could even prescribe “who should do what, when and under which conditions” (Schulz, 1998: 847), limiting decision makers’ autonomy. Therefore, organizational tools enable the firm to address recurring problems in a pre-programmed way, leading to inertia in terms of the application of predefined solutions and increasing organizational inertia. In this respect, for instance, Benner and Tushman (2002) find that ISO 9000 quality program certifications—which include a vast number of written organizational procedures—favor the selection of incremental innovations that are similar to what a firm has done and is doing at the expense of exploratory ones. In our empirical context, knowledge codification could thus reinforce the pattern set by past decisional persistence. If the tools resulting from knowledge codification function as behavioral-control mechanisms, then knowledge codification should strengthen the impact of past integration experience on the probability of integrating a newly acquired target. In contrast, the learning process that occurs over time through codification should have the opposite impact. Given that knowledge codification produces a crisper understanding of the causal linkages between decisions and outcomes, then it should weaken the impact of past decisional persistence on the probability of integrating a newly acquired target. We propose that these two opposing effects can be reconciled and provide a non-monotonic moderating effect of knowledge codification because they emerge at different speeds, and therefore, one effect can prevail over the other at different levels of codification. In particular, because it emerges immediately, the behavioral-control effect of knowledge codification may trump the positive role of learning and allow knowledge codification to increase the persistence in decision-making caused by prior experience—rather than reduce it—until sufficient levels of knowledge codification accumulate. At this point, the effect of learning prevails. Specifically, applying this logic to our empirical context implies that knowledge codification reinforces the inertial effect of integration experience at low levels of codification but reverses it at high levels. The tools that result from low levels of knowledge codification could induce a firm to persist along its decision pattern, as they function as behavioral-control mechanisms before they begin producing a deep understanding of cause-and-effect relationships. In contrast, as the level of codification rises above a certain threshold, the firm becomes able to handle the focal acquisition in a more flexible way, which allows for decision variation according to the specificities of each deal. Formally, we hypothesize the following: H2. As an acquiring organization increases its level of knowledge codification related to integration, the positive impact of the integration experience on the probability of integrating a newly acquired target strengthens and then weakens. Note that whenever both positive and negative effects are at play, the relationship’s precise shape is fundamentally an empirical issue (see Haans et al., 2016 for a review). For instance, some arguments seem to suggest that the inertial effect generated by increasing levels of knowledge codification could prevail over the positive effects of learning. In our context, we expect that the positive effects of codified tools will prevail at high levels of knowledge codification. A high number of codified tools—which can strongly prescribe what to do in the context of simple and repetitive operational activities—cannot precisely and stringently prescribe what to do in our context. Post-acquisition integration decisions are complex, infrequent strategic decisions that cannot be taken mindlessly and automatically based on codified tools alone (Zollo, 2009; Castellaneta and Salvato, 2017). Therefore, higher levels of codified tools should sustain the development of organizational learning without reducing the variance in the observed decision. Thus, a priori, we expect an inverted U-shaped moderation of knowledge codification. 3. Data 3.1 Sample To test our hypotheses, we investigate M&As in the US commercial banking industry between 1985 and 1995.1 Although these data are relatively old, this setting and period offer several benefits. First, the US commercial banking industry started to become very active during that decade, so the conditions were ripe for experience accumulation and knowledge codification and for the emergence of their (positive or negative) effects. Second, evidence suggests that banks displayed significant differences in how they made post-acquisition integration decisions during this period. From 1996 on, however (see DeYoung et al., 2009, for a review), US banks began to pursue mainly efficiency strategies by acquiring inefficient counterparts (Hannan and Pilloff, 2009) or synergy strategies by sharing resources to achieve common objectives (Hagendorff and Keasey, 2009). Both strategies are likely to result in post-acquisition integration and therefore to reduce the likelihood of observing variation in this post-acquisition integration decision. Third, and arguably most importantly, our data set provides rich and fine-grained information on two variables of interest that are rarely available in public data sets and that relate to integration decisions and to the extent of knowledge codification. The research design involved three phases. First, fieldwork was conducted in 12 banks that were active acquirers. This fieldwork sheds light on acquisition practices in the commercial banking industry. Second, 45 decision makers were interviewed to develop a questionnaire-based survey and to check for its measurability and clarity. Third, the survey was administrated among the 250 largest bank-holding companies in the United States in 1996. The survey had two main parts: a profile of acquisition history and a questionnaire for the acquiring bank. In the first portion, respondents were asked to list all acquisitions conducted by their bank and to give basic information about each acquisition, such as asset sizes, the degree of market relatedness, pre-acquisition profitability, and the level of integration. The acquiring-bank questionnaire also elicited information about the ad hoc tools that were developed to manage the acquisition process (e.g., integration manuals, systems conversion manuals, training packages, due-diligence checklists, branch staffing models, and product mapping models). Of the 250 bank-holding companies contacted, 70 did not undertake an acquisition after 1985, and 16 were acquired during the invitation period. Of the remaining 164 banks, 51 responded to the survey, resulting in a 31.7% response rate. To check whether the respondent banks were different from the contacted ones, a standard means comparison test was performed. Responding organizations did not differ from the original set of 250 bank-holding companies in terms of their return on assets (ROA), equity, or efficiency ratios, although they tended to be larger in terms of total assets (P < 0.05). The response rate also appeared to be satisfactory, given the seniority of the respondents and the complexity of the survey. This response rate could also have resulted due to the salience of this topic to industry participants and the in-depth pretests of the survey instrument (Groves et al., 1992). The respondents, who were the most knowledgeable persons at each bank, were identified through telephone calls prior to the mailing and included managers responsible for corporate development or the M&A group (25 cases), coordinators of post-acquisition integration processes (14 cases), CFOs (9 cases), and CEOs (3 cases). Completing the questionnaire offered these respondents an opportunity to benchmark their acquisition practices with those of other firms in the industry. Moreover, the respondents received assurances that their individual responses would remain confidential. Four bank-holding companies were excluded from the analysis because of incomplete responses, and 16 more were excluded because of missing values. The final sample contains data on 31 acquiring banks that completed 408 acquisitions during the period under observation. Due to missing values for same variables, different subsamples are used depending on the variables included in the model. 3.2 Dependent variable 3.2.1 Integration At the core of our study lies the critical strategic decision on the integration of an acquired target’s structures. To measure whether a target was integrated, we use a dummy variable that was equal to 1 if “all systems, procedures, and products (of the two firms) were completely integrated” and equal to 0 in the remaining three cases (if “many but not all systems, procedures, and products were aligned or centralized”; “if only selected systems, procedures, or products were aligned or centralized”; or if “few or no features were aligned or centralized”).2 This measure considers the fact that when a target is not fully integrated, the parent has two options: (i) to centralize some of the target’s activities or (ii) to simply align the activities between the parent and the target, ensuring that these activities are run similarly in both companies. 3.3. Explanatory and control variables 3.3.1 Integration experience Consistent with the previous literature, we hypothesized that as an organization becomes more experienced with a particular decision, it becomes more likely to repeat that decision (Amburgey et al., 1993; Gulati, 1995; Shaver et al., 1997; Dimov et al., 2012). As we apply this theory in the M&A context and, more precisely, considering post-M&A integration decisions, we measure (past) decisional persistence as (the log of) the number of past acquisitions that were integrated. This measure captures the experience that a firm has with a particular decision. 3.3.2 Knowledge codification The measure of knowledge codification is based on the number of tools developed by the acquiring firm at the time of the focal acquisition. These tools reflect the various parts of the acquisition process, such as financial evaluation, due diligence, conversion of information systems, human resource integration, and sales/product integration. The number of tools ranges from 1 to 11 and refers to the due-diligence checklist and to separate manuals on: financial evaluation spreadsheets, due diligence checklist, due diligence manual, info systems conversion manual, info systems training manual, affiliation/integration manual, branch staffing models, training/self-training packages, products training manual, product mapping models, and project management packages (Zollo and Singh, 2004; Zollo, 2009). Consistent with the feedback received in the fieldwork, this list also contains tools—such as the due diligence checklist—that, while not directly linked to the integration process, help in its planning. This measure thereby approximates the acquiring bank’s development of integration practices at the time of the focal acquisition. As with integration experience, we used a log transformation.3 3.3.3 Control variables First, we control for the relatedness between the acquiring firm and the target. Higher market relatedness between the acquiring firm and the target might increase the likelihood of integration because more functions could overlap in their extension and routines. The bank industry clearly distinguishes between “in-market” (horizontal) and “out-market” (market extension) acquisitions; the survey probes the categorization of each transaction listed in the acquisition history profile. Thus, resource relatedness is coded as 1 if the acquisition is in-market and 0 if it is out-market. Second, we control for target quality. The quality of the acquired firm can affect the decision to integrate it such that acquiring firms are less likely to integrate a high-quality target. To measure the pre-acquisition quality of the acquired firm’s resource endowments, we use the following scale: −2 (the acquired institution was bankrupt), −1 (it was a poor performer), 0 (it was an average performer), +1 (it was a good performer), and +2 (it was an outstanding performer). Third, we control for relative acquisition size. The acquired firm’s size relative to that of the acquiring bank (based on total assets) can be stated as a percentage (Datta, 1991). Fourth, we control for parallel acquisitions, that is, the number of acquisitions the acquiring firm manages simultaneously might influence post-acquisition strategies, including integration. A firm that has carried out several acquisitions might lack the energy, time, or resources (Castellaneta and Zollo, 2015) to pursue deep organizational integration. However, more acquisitions might signal the firm’s rapid growth strategy, which can be attained only through strict integration. We account for this effect by controlling for the number of acquisitions concluded by the acquiring firm in the same year as the focal acquisition. Finally, we also control for acquisition experience, computed as the number of acquisitions completed by the acquiring firm before the focal acquisitions (irrespective of whether they were integrated).4 Note that acquisition experience and codification do not necessarily covary: some firms carry out several acquisitions without conducting any knowledge codification, while other firms begin knowledge codification at a very early stage in their acquisition pattern. 4. Analyses and results We study the effect of past decisional experience, knowledge codification, and their interaction on the probability of integrating a target after an M&A deal. Given the binary nature of the dependent variable, we use as the main model a logistic regression with robust standard errors, which accounts for heteroscedasticity in the error distribution. Table 1 reports descriptive statistics and a correlation matrix for the variables used in our hypothesis tests. Table 2 presents the results of the logit model, where explanatory variables are inserted gradually. Table 1. Descriptive statistics and correlation matrix Variables Mean SD 1 2 3 4 5 6 7 8 9 1. Integration 0.718 0.450 1 2. Integration experience 5.027 6.832 0.378*** 1 3. Acquisition experience 11.476 10.162 0.135** 0.637*** 1 4. Relatedness 0.623 0.485 0.417*** 0.221*** 0.168*** 1 5. Parallel acquisitions 3.630 2.826 0.205*** 0.658*** 0.507*** 0.135** 1 6. Target quality −0.0234 1.076 −0.260*** −0.0454 0.0349 −0.223*** 0.0537 1 7. Relative acquisition size 1.057 1.665 −0.0385 0.0455 0.0565 −0.171*** −0.0577 0.0820 1 8. ROA 0.007 0.376 0.239*** 0.128* 0.0115 0.0961 0.195*** −0.120* 0.0789 1 9. Deviation 0.281 0.251 −0.606*** −0.424*** −0.267*** −0.459*** −0.435*** 0.252*** 0.124* −0.263*** 1 10. PoIAs 0.432 0.390 0.516*** 0.612*** 0.0193 0.165*** 0.172*** −0.0901 0.0824 0.135* −0.256*** 1 11. Knowledge codification 4.854 3.677 0.0396 0.328*** 0.459*** 0.0272 0.367*** 0.179*** 0.0766 0.0998 −0.141** 0.0595 Variables Mean SD 1 2 3 4 5 6 7 8 9 1. Integration 0.718 0.450 1 2. Integration experience 5.027 6.832 0.378*** 1 3. Acquisition experience 11.476 10.162 0.135** 0.637*** 1 4. Relatedness 0.623 0.485 0.417*** 0.221*** 0.168*** 1 5. Parallel acquisitions 3.630 2.826 0.205*** 0.658*** 0.507*** 0.135** 1 6. Target quality −0.0234 1.076 −0.260*** −0.0454 0.0349 −0.223*** 0.0537 1 7. Relative acquisition size 1.057 1.665 −0.0385 0.0455 0.0565 −0.171*** −0.0577 0.0820 1 8. ROA 0.007 0.376 0.239*** 0.128* 0.0115 0.0961 0.195*** −0.120* 0.0789 1 9. Deviation 0.281 0.251 −0.606*** −0.424*** −0.267*** −0.459*** −0.435*** 0.252*** 0.124* −0.263*** 1 10. PoIAs 0.432 0.390 0.516*** 0.612*** 0.0193 0.165*** 0.172*** −0.0901 0.0824 0.135* −0.256*** 1 11. Knowledge codification 4.854 3.677 0.0396 0.328*** 0.459*** 0.0272 0.367*** 0.179*** 0.0766 0.0998 −0.141** 0.0595 Note: * P < 0.05; ** P < 0.01; *** P < 0.001. Table 1. Descriptive statistics and correlation matrix Variables Mean SD 1 2 3 4 5 6 7 8 9 1. Integration 0.718 0.450 1 2. Integration experience 5.027 6.832 0.378*** 1 3. Acquisition experience 11.476 10.162 0.135** 0.637*** 1 4. Relatedness 0.623 0.485 0.417*** 0.221*** 0.168*** 1 5. Parallel acquisitions 3.630 2.826 0.205*** 0.658*** 0.507*** 0.135** 1 6. Target quality −0.0234 1.076 −0.260*** −0.0454 0.0349 −0.223*** 0.0537 1 7. Relative acquisition size 1.057 1.665 −0.0385 0.0455 0.0565 −0.171*** −0.0577 0.0820 1 8. ROA 0.007 0.376 0.239*** 0.128* 0.0115 0.0961 0.195*** −0.120* 0.0789 1 9. Deviation 0.281 0.251 −0.606*** −0.424*** −0.267*** −0.459*** −0.435*** 0.252*** 0.124* −0.263*** 1 10. PoIAs 0.432 0.390 0.516*** 0.612*** 0.0193 0.165*** 0.172*** −0.0901 0.0824 0.135* −0.256*** 1 11. Knowledge codification 4.854 3.677 0.0396 0.328*** 0.459*** 0.0272 0.367*** 0.179*** 0.0766 0.0998 −0.141** 0.0595 Variables Mean SD 1 2 3 4 5 6 7 8 9 1. Integration 0.718 0.450 1 2. Integration experience 5.027 6.832 0.378*** 1 3. Acquisition experience 11.476 10.162 0.135** 0.637*** 1 4. Relatedness 0.623 0.485 0.417*** 0.221*** 0.168*** 1 5. Parallel acquisitions 3.630 2.826 0.205*** 0.658*** 0.507*** 0.135** 1 6. Target quality −0.0234 1.076 −0.260*** −0.0454 0.0349 −0.223*** 0.0537 1 7. Relative acquisition size 1.057 1.665 −0.0385 0.0455 0.0565 −0.171*** −0.0577 0.0820 1 8. ROA 0.007 0.376 0.239*** 0.128* 0.0115 0.0961 0.195*** −0.120* 0.0789 1 9. Deviation 0.281 0.251 −0.606*** −0.424*** −0.267*** −0.459*** −0.435*** 0.252*** 0.124* −0.263*** 1 10. PoIAs 0.432 0.390 0.516*** 0.612*** 0.0193 0.165*** 0.172*** −0.0901 0.0824 0.135* −0.256*** 1 11. Knowledge codification 4.854 3.677 0.0396 0.328*** 0.459*** 0.0272 0.367*** 0.179*** 0.0766 0.0998 −0.141** 0.0595 Note: * P < 0.05; ** P < 0.01; *** P < 0.001. Table 2. Logit predicting integration probability with integration experience Variables (1) (2) (3) (4) (5) (6) Integration Integration Integration Integration Integration Integration Integration experience 0.952*** 1.115*** 1.106*** 0.962*** 1.188*** (0.150) (0.170) (0.171) (0.295) (0.277) Knowledge codification 0.209 0.388 0.134 −0.079 (0.166) (0.264) (0.188) (0.328) Integration experience × knowledge codification 0.105 0.670** (0.167) (0.273) Knowledge codification ^2 −0.158 0.085 (0.177) (0.199) Integration experience × knowledge codification ^2 −0.330** (0.137) Acquisition experience −0.007 −0.047** −0.041 −0.034 −0.039 −0.036 (0.015) (0.022) (0.025) (0.026) (0.026) (0.026) Relatedness 1.941*** 2.140*** 2.378*** 2.398*** 2.363*** 2.545*** (0.291) (0.355) (0.396) (0.398) (0.396) (0.412) Parallel acquisitions 0.391*** 0.369*** 0.210** 0.236** 0.208** 0.255** (0.077) (0.083) (0.095) (0.100) (0.097) (0.103) Target quality −0.459*** −0.358* −0.484** −0.470** −0.488** −0.532*** (0.138) (0.185) (0.206) (0.201) (0.204) (0.204) Relative acquisition size 0.069 0.021 −0.109 −0.103 −0.119 −0.112 (0.070) (0.071) (0.095) (0.097) (0.099) (0.102) Constant −1.254*** −1.574*** −1.612*** −1.564*** −1.503*** −1.716*** (0.344) (0.400) (0.445) (0.465) (0.487) (0.531) Observations 384 384 358 358 358 358 Pseudo R-squared 0.255 0.398 0.451 0.453 0.452 0.470 Log likelihood −161.592 −130.664 −114.345 −113.950 −114.066 −110.317 Variables (1) (2) (3) (4) (5) (6) Integration Integration Integration Integration Integration Integration Integration experience 0.952*** 1.115*** 1.106*** 0.962*** 1.188*** (0.150) (0.170) (0.171) (0.295) (0.277) Knowledge codification 0.209 0.388 0.134 −0.079 (0.166) (0.264) (0.188) (0.328) Integration experience × knowledge codification 0.105 0.670** (0.167) (0.273) Knowledge codification ^2 −0.158 0.085 (0.177) (0.199) Integration experience × knowledge codification ^2 −0.330** (0.137) Acquisition experience −0.007 −0.047** −0.041 −0.034 −0.039 −0.036 (0.015) (0.022) (0.025) (0.026) (0.026) (0.026) Relatedness 1.941*** 2.140*** 2.378*** 2.398*** 2.363*** 2.545*** (0.291) (0.355) (0.396) (0.398) (0.396) (0.412) Parallel acquisitions 0.391*** 0.369*** 0.210** 0.236** 0.208** 0.255** (0.077) (0.083) (0.095) (0.100) (0.097) (0.103) Target quality −0.459*** −0.358* −0.484** −0.470** −0.488** −0.532*** (0.138) (0.185) (0.206) (0.201) (0.204) (0.204) Relative acquisition size 0.069 0.021 −0.109 −0.103 −0.119 −0.112 (0.070) (0.071) (0.095) (0.097) (0.099) (0.102) Constant −1.254*** −1.574*** −1.612*** −1.564*** −1.503*** −1.716*** (0.344) (0.400) (0.445) (0.465) (0.487) (0.531) Observations 384 384 358 358 358 358 Pseudo R-squared 0.255 0.398 0.451 0.453 0.452 0.470 Log likelihood −161.592 −130.664 −114.345 −113.950 −114.066 −110.317 Note: Robust standard errors in parentheses. * P < 0.1; ** P < 0.05; *** P < 0.01. Table 2. Logit predicting integration probability with integration experience Variables (1) (2) (3) (4) (5) (6) Integration Integration Integration Integration Integration Integration Integration experience 0.952*** 1.115*** 1.106*** 0.962*** 1.188*** (0.150) (0.170) (0.171) (0.295) (0.277) Knowledge codification 0.209 0.388 0.134 −0.079 (0.166) (0.264) (0.188) (0.328) Integration experience × knowledge codification 0.105 0.670** (0.167) (0.273) Knowledge codification ^2 −0.158 0.085 (0.177) (0.199) Integration experience × knowledge codification ^2 −0.330** (0.137) Acquisition experience −0.007 −0.047** −0.041 −0.034 −0.039 −0.036 (0.015) (0.022) (0.025) (0.026) (0.026) (0.026) Relatedness 1.941*** 2.140*** 2.378*** 2.398*** 2.363*** 2.545*** (0.291) (0.355) (0.396) (0.398) (0.396) (0.412) Parallel acquisitions 0.391*** 0.369*** 0.210** 0.236** 0.208** 0.255** (0.077) (0.083) (0.095) (0.100) (0.097) (0.103) Target quality −0.459*** −0.358* −0.484** −0.470** −0.488** −0.532*** (0.138) (0.185) (0.206) (0.201) (0.204) (0.204) Relative acquisition size 0.069 0.021 −0.109 −0.103 −0.119 −0.112 (0.070) (0.071) (0.095) (0.097) (0.099) (0.102) Constant −1.254*** −1.574*** −1.612*** −1.564*** −1.503*** −1.716*** (0.344) (0.400) (0.445) (0.465) (0.487) (0.531) Observations 384 384 358 358 358 358 Pseudo R-squared 0.255 0.398 0.451 0.453 0.452 0.470 Log likelihood −161.592 −130.664 −114.345 −113.950 −114.066 −110.317 Variables (1) (2) (3) (4) (5) (6) Integration Integration Integration Integration Integration Integration Integration experience 0.952*** 1.115*** 1.106*** 0.962*** 1.188*** (0.150) (0.170) (0.171) (0.295) (0.277) Knowledge codification 0.209 0.388 0.134 −0.079 (0.166) (0.264) (0.188) (0.328) Integration experience × knowledge codification 0.105 0.670** (0.167) (0.273) Knowledge codification ^2 −0.158 0.085 (0.177) (0.199) Integration experience × knowledge codification ^2 −0.330** (0.137) Acquisition experience −0.007 −0.047** −0.041 −0.034 −0.039 −0.036 (0.015) (0.022) (0.025) (0.026) (0.026) (0.026) Relatedness 1.941*** 2.140*** 2.378*** 2.398*** 2.363*** 2.545*** (0.291) (0.355) (0.396) (0.398) (0.396) (0.412) Parallel acquisitions 0.391*** 0.369*** 0.210** 0.236** 0.208** 0.255** (0.077) (0.083) (0.095) (0.100) (0.097) (0.103) Target quality −0.459*** −0.358* −0.484** −0.470** −0.488** −0.532*** (0.138) (0.185) (0.206) (0.201) (0.204) (0.204) Relative acquisition size 0.069 0.021 −0.109 −0.103 −0.119 −0.112 (0.070) (0.071) (0.095) (0.097) (0.099) (0.102) Constant −1.254*** −1.574*** −1.612*** −1.564*** −1.503*** −1.716*** (0.344) (0.400) (0.445) (0.465) (0.487) (0.531) Observations 384 384 358 358 358 358 Pseudo R-squared 0.255 0.398 0.451 0.453 0.452 0.470 Log likelihood −161.592 −130.664 −114.345 −113.950 −114.066 −110.317 Note: Robust standard errors in parentheses. * P < 0.1; ** P < 0.05; *** P < 0.01. H1 posited that as integration experience increases, the probability of integrating a newly acquired target increases. The positive and significant coefficient of integration experience (Model 6 of Table 2) provides support for H1 (P < 0.01). However, we also claimed that integration experience likely produces persistence in decision-making not because of learning but rather because of inertia. However, the fact that a firm repeats the same decision over time does not automatically imply that the decision is harmful for the organization. In contrast, one might argue that a decision is repeated over time because of increasing returns created by the feedback between experience and competence. Hence, the acquiring firm might continue to conduct the same “type of deal” for which it is advisable to continue integrating because of the firm’s knowledge of and capabilities with respect to the integration process. If this is the case, then there is no need to diminish persistence in decision-making. Because we cannot directly observe whether superior capabilities or inertia drive the persistence effect, we resort to preliminary indirect empirical evidence according to the following steps. Step 1: Predicting the integration decision. In the first step, we predict—through a logit model—the expected integration decision for each deal (i.e., the probability, pint, that two merging firms are actually integrated) based on four relevant predictors of the integration decision that are consistently suggested by the related literature. In other words, we first estimate a logit model in which the dependent variable is the integration decision and the covariates are relatedness, target quality, relative acquisition size, and parallel acquisitions. Given the result of this estimation, for each deal, we then predict the probability pint that the target is integrated, conditional on the specific values that the four covariates assume in the specific deal. In essence, the pint variable thus captures how an acquiring firm in our sample would handle, on average, its post-acquisition integration decision, given the actual value of the four predictors. By construction, pint is bounded by 0 and 1. Step 2: Calculating the deviation. In the second step, for each deal, we create a Deviation variable such that Deviation = | Actual integration decision – pint |. If the predicted pint is low (i.e., the four variables suggest that the deal was unlikely to be integrated) but the target was eventually integrated, the value of Deviation is relatively high. If the deal was integrated and pint is high, then the value of Deviation is low. Similarly, if the deal was not actually integrated and pint is low, Deviation is negative but has a low absolute value, whereas if pint is high, Deviation is negative and has a high absolute value. Step 3: Deviation and performance. In the third step, we calculate a measure of deal performance and regress it on the absolute value of Deviation and the control variables used in the first step. Following Zollo and Singh (2004), we measure acquisition performance as the difference between the ROA of the acquiring bank 3 years after the acquisition (which considers the entire acquiring bank, including the acquired target) versus 1 year before (which considers only the acquiring bank).5 This step checks whether a decision regarding organizational integration—that “deviates” from what deal characteristics indicate an average sample firm would do—has a positive or a negative effect on performance. As the previous regressions have highlighted, this possible “deviation” is due to persistence in decision-making.6 However, persistence due to increasing returns and superior capabilities would be mirrored in a positive effect of deviation on performance. Conversely, if persistence were related to organizational inertia, we would expect the effect of deviation to be negative. The results of this last linear regression, which is reported in Model 2 of Table 3, show that the parameter estimate of the absolute value of Deviation is negative and significant (P < 0.1). Therefore, with all other things being equal, as a firm deviates more from the “predicted” decision (i.e., deviates from what the firms in our sample would have done on average in that specific situation), its deal performance decreases. Hence, this result suggests that persistence in decision-making is associated with negative results in terms of performance. Clearly, these results are only indicative and correlational in nature, as our data structure does not allow us to properly disentangle inertia from learning; still, for this sample, preliminary evidence is consistent with the notion that persistence and organizational inertia are associated. Table 3. The effect of relatedness, parallel acquisitions, target quality, and relative acquisition size on integration probability (Model 1); the effect of deviation (absolute value) on acquisition performance measured as ROA (Model 2) Variables (1) (2) Integration Acquisition performance (ROA) Deviation (absolute value) −0.275* (0.117) Relatedness 1.912*** 0.029 (0.276) (0.058) Parallel acquisitions 0.375*** 0.015** (0.071) (0.005) Target quality −0.456*** −0.030 (0.133) (0.024) Relative acquisition size 0.078 0.025 (0.068) (0.015) Constant −1.275*** −0.043 (0.324) (0.072) Observations 408 259 Log likelihood −175.332 0.095 Variables (1) (2) Integration Acquisition performance (ROA) Deviation (absolute value) −0.275* (0.117) Relatedness 1.912*** 0.029 (0.276) (0.058) Parallel acquisitions 0.375*** 0.015** (0.071) (0.005) Target quality −0.456*** −0.030 (0.133) (0.024) Relative acquisition size 0.078 0.025 (0.068) (0.015) Constant −1.275*** −0.043 (0.324) (0.072) Observations 408 259 Log likelihood −175.332 0.095 Note: Robust standard errors in parentheses. * P < 0.1; ** P < 0.05; *** P < 0.01. Table 3. The effect of relatedness, parallel acquisitions, target quality, and relative acquisition size on integration probability (Model 1); the effect of deviation (absolute value) on acquisition performance measured as ROA (Model 2) Variables (1) (2) Integration Acquisition performance (ROA) Deviation (absolute value) −0.275* (0.117) Relatedness 1.912*** 0.029 (0.276) (0.058) Parallel acquisitions 0.375*** 0.015** (0.071) (0.005) Target quality −0.456*** −0.030 (0.133) (0.024) Relative acquisition size 0.078 0.025 (0.068) (0.015) Constant −1.275*** −0.043 (0.324) (0.072) Observations 408 259 Log likelihood −175.332 0.095 Variables (1) (2) Integration Acquisition performance (ROA) Deviation (absolute value) −0.275* (0.117) Relatedness 1.912*** 0.029 (0.276) (0.058) Parallel acquisitions 0.375*** 0.015** (0.071) (0.005) Target quality −0.456*** −0.030 (0.133) (0.024) Relative acquisition size 0.078 0.025 (0.068) (0.015) Constant −1.275*** −0.043 (0.324) (0.072) Observations 408 259 Log likelihood −175.332 0.095 Note: Robust standard errors in parentheses. * P < 0.1; ** P < 0.05; *** P < 0.01. H2 suggests that as an acquiring organization increases its level of knowledge codification related to integration, integration experience’s positive impact on the probability of integrating a newly acquired target strengthens and then weakens. That is, we propose a curvilinear moderating effect. The interaction term between integration experience and knowledge codification is positive and significant (P < 0.05), and that between integration experience and knowledge codification squared is negative and significant (P < 0.05) in Model 6 of Table 2. Overall, these findings support H2. The previous literature has proposed that persistence is a result of the amount of accumulated experience (Amburgey and Miner, 1992). The core idea is that frequent repetitions of a decision cause decision-making persistence and path dependence (Katz and Shapiro, 1985; Levinthal and March, 1993). However, in principle, path dependence can be created at any level of experience. The only condition necessary for the emergence of path-dependent behaviors is that the organization “fall into the trap” of certain constraints, such as routines, cognitive schemata, or structures, which, despite being influenced by repetition, can emerge regardless of the quantity of experience accumulated. Hence, as a robustness check, we operationalize past decision persistence as the percentage of integrated acquisitions (PoIAs) (i.e., the share of integrated deals among all the acquiring firm’s previous deals). Model 6 in Table 4 shows that our results are robust to the use of this operationalization of past persistence. Table 4. Logit predicting integration probability with the PoIAs Variables (1) (2) (3) (4) (5) (6) Integration Integration Integration Integration Integration Integration PoIAs 3.467*** 3.960*** 3.953*** 2.413*** 3.805*** (0.520) (0.563) (0.564) (0.714) (0.995) Knowledge codification 0.117 0.426 −0.475* −0.806* (0.183) (0.297) (0.253) (0.471) PoIA × knowledge codification 1.454*** 2.826*** (0.492) (0.860) Knowledge codification ^ 2 −0.255 0.227 (0.174) (0.269) PoIA × knowledge codification ^ 2 −1.107** (0.540) Acquisition experience −0.008 0.012 0.029 0.039* 0.039* 0.039* (0.014) (0.017) (0.019) (0.021) (0.021) (0.021) Relatedness 1.927*** 2.086*** 2.304*** 2.326*** 2.439*** 2.648*** (0.279) (0.355) (0.386) (0.385) (0.402) (0.430) Parallel acquisitions 0.389*** 0.404*** 0.265*** 0.314*** 0.241** 0.283** (0.075) (0.083) (0.102) (0.110) (0.114) (0.119) Target quality −0.452*** −0.496*** −0.636*** −0.610*** −0.612*** −0.631*** (0.133) (0.177) (0.195) (0.190) (0.193) (0.192) Relative acquisition size 0.081 −0.025 −0.140 −0.133 −0.149 −0.130 (0.068) (0.074) (0.117) (0.120) (0.121) (0.123) Constant −1.254*** −2.889*** −3.069*** −3.020*** −2.479*** −2.946*** (0.327) (0.503) (0.549) (0.584) (0.614) (0.691) Observations 408 384 358 358 358 358 Pseudo R-squared 0.254 0.407 0.453 0.458 0.476 0.491 Log likelihood −175.211 −128.607 −113.885 −112.858 −108.968 −106.028 Variables (1) (2) (3) (4) (5) (6) Integration Integration Integration Integration Integration Integration PoIAs 3.467*** 3.960*** 3.953*** 2.413*** 3.805*** (0.520) (0.563) (0.564) (0.714) (0.995) Knowledge codification 0.117 0.426 −0.475* −0.806* (0.183) (0.297) (0.253) (0.471) PoIA × knowledge codification 1.454*** 2.826*** (0.492) (0.860) Knowledge codification ^ 2 −0.255 0.227 (0.174) (0.269) PoIA × knowledge codification ^ 2 −1.107** (0.540) Acquisition experience −0.008 0.012 0.029 0.039* 0.039* 0.039* (0.014) (0.017) (0.019) (0.021) (0.021) (0.021) Relatedness 1.927*** 2.086*** 2.304*** 2.326*** 2.439*** 2.648*** (0.279) (0.355) (0.386) (0.385) (0.402) (0.430) Parallel acquisitions 0.389*** 0.404*** 0.265*** 0.314*** 0.241** 0.283** (0.075) (0.083) (0.102) (0.110) (0.114) (0.119) Target quality −0.452*** −0.496*** −0.636*** −0.610*** −0.612*** −0.631*** (0.133) (0.177) (0.195) (0.190) (0.193) (0.192) Relative acquisition size 0.081 −0.025 −0.140 −0.133 −0.149 −0.130 (0.068) (0.074) (0.117) (0.120) (0.121) (0.123) Constant −1.254*** −2.889*** −3.069*** −3.020*** −2.479*** −2.946*** (0.327) (0.503) (0.549) (0.584) (0.614) (0.691) Observations 408 384 358 358 358 358 Pseudo R-squared 0.254 0.407 0.453 0.458 0.476 0.491 Log likelihood −175.211 −128.607 −113.885 −112.858 −108.968 −106.028 Note: Robust standard errors in parentheses. * P < 0.1; ** P < 0.05; *** P < 0.01. Table 4. Logit predicting integration probability with the PoIAs Variables (1) (2) (3) (4) (5) (6) Integration Integration Integration Integration Integration Integration PoIAs 3.467*** 3.960*** 3.953*** 2.413*** 3.805*** (0.520) (0.563) (0.564) (0.714) (0.995) Knowledge codification 0.117 0.426 −0.475* −0.806* (0.183) (0.297) (0.253) (0.471) PoIA × knowledge codification 1.454*** 2.826*** (0.492) (0.860) Knowledge codification ^ 2 −0.255 0.227 (0.174) (0.269) PoIA × knowledge codification ^ 2 −1.107** (0.540) Acquisition experience −0.008 0.012 0.029 0.039* 0.039* 0.039* (0.014) (0.017) (0.019) (0.021) (0.021) (0.021) Relatedness 1.927*** 2.086*** 2.304*** 2.326*** 2.439*** 2.648*** (0.279) (0.355) (0.386) (0.385) (0.402) (0.430) Parallel acquisitions 0.389*** 0.404*** 0.265*** 0.314*** 0.241** 0.283** (0.075) (0.083) (0.102) (0.110) (0.114) (0.119) Target quality −0.452*** −0.496*** −0.636*** −0.610*** −0.612*** −0.631*** (0.133) (0.177) (0.195) (0.190) (0.193) (0.192) Relative acquisition size 0.081 −0.025 −0.140 −0.133 −0.149 −0.130 (0.068) (0.074) (0.117) (0.120) (0.121) (0.123) Constant −1.254*** −2.889*** −3.069*** −3.020*** −2.479*** −2.946*** (0.327) (0.503) (0.549) (0.584) (0.614) (0.691) Observations 408 384 358 358 358 358 Pseudo R-squared 0.254 0.407 0.453 0.458 0.476 0.491 Log likelihood −175.211 −128.607 −113.885 −112.858 −108.968 −106.028 Variables (1) (2) (3) (4) (5) (6) Integration Integration Integration Integration Integration Integration PoIAs 3.467*** 3.960*** 3.953*** 2.413*** 3.805*** (0.520) (0.563) (0.564) (0.714) (0.995) Knowledge codification 0.117 0.426 −0.475* −0.806* (0.183) (0.297) (0.253) (0.471) PoIA × knowledge codification 1.454*** 2.826*** (0.492) (0.860) Knowledge codification ^ 2 −0.255 0.227 (0.174) (0.269) PoIA × knowledge codification ^ 2 −1.107** (0.540) Acquisition experience −0.008 0.012 0.029 0.039* 0.039* 0.039* (0.014) (0.017) (0.019) (0.021) (0.021) (0.021) Relatedness 1.927*** 2.086*** 2.304*** 2.326*** 2.439*** 2.648*** (0.279) (0.355) (0.386) (0.385) (0.402) (0.430) Parallel acquisitions 0.389*** 0.404*** 0.265*** 0.314*** 0.241** 0.283** (0.075) (0.083) (0.102) (0.110) (0.114) (0.119) Target quality −0.452*** −0.496*** −0.636*** −0.610*** −0.612*** −0.631*** (0.133) (0.177) (0.195) (0.190) (0.193) (0.192) Relative acquisition size 0.081 −0.025 −0.140 −0.133 −0.149 −0.130 (0.068) (0.074) (0.117) (0.120) (0.121) (0.123) Constant −1.254*** −2.889*** −3.069*** −3.020*** −2.479*** −2.946*** (0.327) (0.503) (0.549) (0.584) (0.614) (0.691) Observations 408 384 358 358 358 358 Pseudo R-squared 0.254 0.407 0.453 0.458 0.476 0.491 Log likelihood −175.211 −128.607 −113.885 −112.858 −108.968 −106.028 Note: Robust standard errors in parentheses. * P < 0.1; ** P < 0.05; *** P < 0.01. The previous results are based on interaction terms in logistic regressions. However, because logit models are nonlinear models, the parameter estimates of the interaction terms should be carefully interpreted. To address this issue, we also adopt a linear probability model approach, which allows us to obtain a direct and simple estimate of the interaction effect. Because the predicted probabilities lie between 0 and 1, the linear probability model does not produce biased estimates (Horrace and Oaxaca, 2006). The results reported in Tables 5 and 6 show that our results are also robust to this specification. Moreover, based on Model 6 in Table 5, we graph the interaction effect between knowledge codification and integration experience (Figure 1). Table 5. Linear probability model predicting integration probability with integration experience Variables (1) (2) (3) (4) (5) (6) Integration Integration Integration Integration Integration Integration Integration experience 0.138*** 0.164*** 0.163*** 0.166*** 0.188*** (0.018) (0.019) (0.019) (0.038) (0.037) Knowledge codification 0.027 0.042 0.028 −0.007 (0.020) (0.034) (0.026) (0.047) Integration experience × Knowledge codification −0.001 0.059 (0.020) (0.036) Knowledge codification ^ 2 −0.012 0.021 (0.021) (0.026) Integration experience × Knowledge codification ^ 2 −0.035** (0.017) Acquisition experience −0.002 −0.009*** −0.007*** −0.007*** −0.007*** −0.008*** (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) Relatedness 0.339*** 0.300*** 0.313*** 0.313*** 0.313*** 0.320*** (0.050) (0.047) (0.049) (0.049) (0.049) (0.049) Parallel acquisitions 0.034*** 0.020*** 0.002 0.003 0.002 0.005 (0.006) (0.006) (0.006) (0.006) (0.007) (0.007) Target quality −0.066*** −0.044** −0.060*** −0.058*** −0.059*** −0.060*** (0.017) (0.018) (0.019) (0.018) (0.019) (0.019) Relative acquisition size 0.012 0.006 −0.008 −0.008 −0.008 −0.008 (0.012) (0.010) (0.014) (0.014) (0.014) (0.014) Constant 0.389*** 0.401*** 0.369*** 0.379*** 0.367*** 0.349*** (0.050) (0.046) (0.051) (0.055) (0.060) (0.063) Observations 384 384 358 358 358 358 R-squared 0.252 0.384 0.452 0.453 0.452 0.462 Variables (1) (2) (3) (4) (5) (6) Integration Integration Integration Integration Integration Integration Integration experience 0.138*** 0.164*** 0.163*** 0.166*** 0.188*** (0.018) (0.019) (0.019) (0.038) (0.037) Knowledge codification 0.027 0.042 0.028 −0.007 (0.020) (0.034) (0.026) (0.047) Integration experience × Knowledge codification −0.001 0.059 (0.020) (0.036) Knowledge codification ^ 2 −0.012 0.021 (0.021) (0.026) Integration experience × Knowledge codification ^ 2 −0.035** (0.017) Acquisition experience −0.002 −0.009*** −0.007*** −0.007*** −0.007*** −0.008*** (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) Relatedness 0.339*** 0.300*** 0.313*** 0.313*** 0.313*** 0.320*** (0.050) (0.047) (0.049) (0.049) (0.049) (0.049) Parallel acquisitions 0.034*** 0.020*** 0.002 0.003 0.002 0.005 (0.006) (0.006) (0.006) (0.006) (0.007) (0.007) Target quality −0.066*** −0.044** −0.060*** −0.058*** −0.059*** −0.060*** (0.017) (0.018) (0.019) (0.018) (0.019) (0.019) Relative acquisition size 0.012 0.006 −0.008 −0.008 −0.008 −0.008 (0.012) (0.010) (0.014) (0.014) (0.014) (0.014) Constant 0.389*** 0.401*** 0.369*** 0.379*** 0.367*** 0.349*** (0.050) (0.046) (0.051) (0.055) (0.060) (0.063) Observations 384 384 358 358 358 358 R-squared 0.252 0.384 0.452 0.453 0.452 0.462 Note: Robust standard errors in parentheses. * P < 0.1; ** P < 0.05; *** P < 0.01. Table 5. Linear probability model predicting integration probability with integration experience Variables (1) (2) (3) (4) (5) (6) Integration Integration Integration Integration Integration Integration Integration experience 0.138*** 0.164*** 0.163*** 0.166*** 0.188*** (0.018) (0.019) (0.019) (0.038) (0.037) Knowledge codification 0.027 0.042 0.028 −0.007 (0.020) (0.034) (0.026) (0.047) Integration experience × Knowledge codification −0.001 0.059 (0.020) (0.036) Knowledge codification ^ 2 −0.012 0.021 (0.021) (0.026) Integration experience × Knowledge codification ^ 2 −0.035** (0.017) Acquisition experience −0.002 −0.009*** −0.007*** −0.007*** −0.007*** −0.008*** (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) Relatedness 0.339*** 0.300*** 0.313*** 0.313*** 0.313*** 0.320*** (0.050) (0.047) (0.049) (0.049) (0.049) (0.049) Parallel acquisitions 0.034*** 0.020*** 0.002 0.003 0.002 0.005 (0.006) (0.006) (0.006) (0.006) (0.007) (0.007) Target quality −0.066*** −0.044** −0.060*** −0.058*** −0.059*** −0.060*** (0.017) (0.018) (0.019) (0.018) (0.019) (0.019) Relative acquisition size 0.012 0.006 −0.008 −0.008 −0.008 −0.008 (0.012) (0.010) (0.014) (0.014) (0.014) (0.014) Constant 0.389*** 0.401*** 0.369*** 0.379*** 0.367*** 0.349*** (0.050) (0.046) (0.051) (0.055) (0.060) (0.063) Observations 384 384 358 358 358 358 R-squared 0.252 0.384 0.452 0.453 0.452 0.462 Variables (1) (2) (3) (4) (5) (6) Integration Integration Integration Integration Integration Integration Integration experience 0.138*** 0.164*** 0.163*** 0.166*** 0.188*** (0.018) (0.019) (0.019) (0.038) (0.037) Knowledge codification 0.027 0.042 0.028 −0.007 (0.020) (0.034) (0.026) (0.047) Integration experience × Knowledge codification −0.001 0.059 (0.020) (0.036) Knowledge codification ^ 2 −0.012 0.021 (0.021) (0.026) Integration experience × Knowledge codification ^ 2 −0.035** (0.017) Acquisition experience −0.002 −0.009*** −0.007*** −0.007*** −0.007*** −0.008*** (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) Relatedness 0.339*** 0.300*** 0.313*** 0.313*** 0.313*** 0.320*** (0.050) (0.047) (0.049) (0.049) (0.049) (0.049) Parallel acquisitions 0.034*** 0.020*** 0.002 0.003 0.002 0.005 (0.006) (0.006) (0.006) (0.006) (0.007) (0.007) Target quality −0.066*** −0.044** −0.060*** −0.058*** −0.059*** −0.060*** (0.017) (0.018) (0.019) (0.018) (0.019) (0.019) Relative acquisition size 0.012 0.006 −0.008 −0.008 −0.008 −0.008 (0.012) (0.010) (0.014) (0.014) (0.014) (0.014) Constant 0.389*** 0.401*** 0.369*** 0.379*** 0.367*** 0.349*** (0.050) (0.046) (0.051) (0.055) (0.060) (0.063) Observations 384 384 358 358 358 358 R-squared 0.252 0.384 0.452 0.453 0.452 0.462 Note: Robust standard errors in parentheses. * P < 0.1; ** P < 0.05; *** P < 0.01. Table 6. Linear probability model predicting integration probability with the PoIAs Variables (1) (2) (3) (4) (5) (6) Integration Integration Integration Integration Integration Integration PoIAs 0.432*** 0.506*** 0.504*** 0.345*** 0.445*** (0.055) (0.059) (0.059) (0.099) (0.104) Knowledge codification 0.016 0.038 −0.044 −0.093 (0.022) (0.035) (0.035) (0.061) PoIA × knowledge codification 0.128** 0.277*** (0.059) (0.100) Knowledge codification ^ 2 −0.018 0.034 (0.021) (0.035) PoIA × knowledge codification ^ 2 −0.109** (0.055) Acquisition experience −0.001 0.000 0.003 0.004* 0.004* 0.004* (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) Relatedness 0.342*** 0.290*** 0.306*** 0.306*** 0.304*** 0.318*** (0.048) (0.047) (0.049) (0.049) (0.049) (0.049) Parallel acquisitions 0.035*** 0.022*** 0.006 0.007 −0.000 0.004 (0.006) (0.006) (0.006) (0.006) (0.007) (0.007) Target quality −0.067*** −0.060*** −0.076*** −0.074*** −0.075*** −0.075*** (0.017) (0.017) (0.018) (0.018) (0.018) (0.017) Relative acquisition size 0.015 0.003 −0.010 −0.010 −0.011 −0.010 (0.011) (0.010) (0.014) (0.014) (0.014) (0.014) Constant 0.378*** 0.239*** 0.186*** 0.202*** 0.267*** 0.223*** (0.048) (0.050) (0.054) (0.058) (0.068) (0.077) Observations 408 384 358 358 358 358 R-squared 0.254 0.373 0.432 0.433 0.442 0.450 Variables (1) (2) (3) (4) (5) (6) Integration Integration Integration Integration Integration Integration PoIAs 0.432*** 0.506*** 0.504*** 0.345*** 0.445*** (0.055) (0.059) (0.059) (0.099) (0.104) Knowledge codification 0.016 0.038 −0.044 −0.093 (0.022) (0.035) (0.035) (0.061) PoIA × knowledge codification 0.128** 0.277*** (0.059) (0.100) Knowledge codification ^ 2 −0.018 0.034 (0.021) (0.035) PoIA × knowledge codification ^ 2 −0.109** (0.055) Acquisition experience −0.001 0.000 0.003 0.004* 0.004* 0.004* (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) Relatedness 0.342*** 0.290*** 0.306*** 0.306*** 0.304*** 0.318*** (0.048) (0.047) (0.049) (0.049) (0.049) (0.049) Parallel acquisitions 0.035*** 0.022*** 0.006 0.007 −0.000 0.004 (0.006) (0.006) (0.006) (0.006) (0.007) (0.007) Target quality −0.067*** −0.060*** −0.076*** −0.074*** −0.075*** −0.075*** (0.017) (0.017) (0.018) (0.018) (0.018) (0.017) Relative acquisition size 0.015 0.003 −0.010 −0.010 −0.011 −0.010 (0.011) (0.010) (0.014) (0.014) (0.014) (0.014) Constant 0.378*** 0.239*** 0.186*** 0.202*** 0.267*** 0.223*** (0.048) (0.050) (0.054) (0.058) (0.068) (0.077) Observations 408 384 358 358 358 358 R-squared 0.254 0.373 0.432 0.433 0.442 0.450 Note: Robust standard errors in parentheses. * P < 0.1; ** P < 0.05; *** P < 0.01. Table 6. Linear probability model predicting integration probability with the PoIAs Variables (1) (2) (3) (4) (5) (6) Integration Integration Integration Integration Integration Integration PoIAs 0.432*** 0.506*** 0.504*** 0.345*** 0.445*** (0.055) (0.059) (0.059) (0.099) (0.104) Knowledge codification 0.016 0.038 −0.044 −0.093 (0.022) (0.035) (0.035) (0.061) PoIA × knowledge codification 0.128** 0.277*** (0.059) (0.100) Knowledge codification ^ 2 −0.018 0.034 (0.021) (0.035) PoIA × knowledge codification ^ 2 −0.109** (0.055) Acquisition experience −0.001 0.000 0.003 0.004* 0.004* 0.004* (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) Relatedness 0.342*** 0.290*** 0.306*** 0.306*** 0.304*** 0.318*** (0.048) (0.047) (0.049) (0.049) (0.049) (0.049) Parallel acquisitions 0.035*** 0.022*** 0.006 0.007 −0.000 0.004 (0.006) (0.006) (0.006) (0.006) (0.007) (0.007) Target quality −0.067*** −0.060*** −0.076*** −0.074*** −0.075*** −0.075*** (0.017) (0.017) (0.018) (0.018) (0.018) (0.017) Relative acquisition size 0.015 0.003 −0.010 −0.010 −0.011 −0.010 (0.011) (0.010) (0.014) (0.014) (0.014) (0.014) Constant 0.378*** 0.239*** 0.186*** 0.202*** 0.267*** 0.223*** (0.048) (0.050) (0.054) (0.058) (0.068) (0.077) Observations 408 384 358 358 358 358 R-squared 0.254 0.373 0.432 0.433 0.442 0.450 Variables (1) (2) (3) (4) (5) (6) Integration Integration Integration Integration Integration Integration PoIAs 0.432*** 0.506*** 0.504*** 0.345*** 0.445*** (0.055) (0.059) (0.059) (0.099) (0.104) Knowledge codification 0.016 0.038 −0.044 −0.093 (0.022) (0.035) (0.035) (0.061) PoIA × knowledge codification 0.128** 0.277*** (0.059) (0.100) Knowledge codification ^ 2 −0.018 0.034 (0.021) (0.035) PoIA × knowledge codification ^ 2 −0.109** (0.055) Acquisition experience −0.001 0.000 0.003 0.004* 0.004* 0.004* (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) Relatedness 0.342*** 0.290*** 0.306*** 0.306*** 0.304*** 0.318*** (0.048) (0.047) (0.049) (0.049) (0.049) (0.049) Parallel acquisitions 0.035*** 0.022*** 0.006 0.007 −0.000 0.004 (0.006) (0.006) (0.006) (0.006) (0.007) (0.007) Target quality −0.067*** −0.060*** −0.076*** −0.074*** −0.075*** −0.075*** (0.017) (0.017) (0.018) (0.018) (0.018) (0.017) Relative acquisition size 0.015 0.003 −0.010 −0.010 −0.011 −0.010 (0.011) (0.010) (0.014) (0.014) (0.014) (0.014) Constant 0.378*** 0.239*** 0.186*** 0.202*** 0.267*** 0.223*** (0.048) (0.050) (0.054) (0.058) (0.068) (0.077) Observations 408 384 358 358 358 358 R-squared 0.254 0.373 0.432 0.433 0.442 0.450 Note: Robust standard errors in parentheses. * P < 0.1; ** P < 0.05; *** P < 0.01. Figure 1. View largeDownload slide Moderating effect of knowledge codification on the relationship between integration experience and integration (based on Model 6 in Table 5). Figure 1. View largeDownload slide Moderating effect of knowledge codification on the relationship between integration experience and integration (based on Model 6 in Table 5). Finally, we study how persistence in acquisition integration is intertwined with momentum in acquisition. In particular, we study whether the persistence in decision-making generated by integration experience increases as acquisition experience increases. Given that both acquisition and integration experience are likely to be subject to routinization processes, they could reinforce each other as they increase more. Using a logistic regression, Model 1 in Table 7 shows that the interaction term between integration experience and acquisition experience is not significant, suggesting that the impact of integration experience on the probability of integrating a newly acquired target does not depend on acquisition experience. We find similar results if we operationalize integration experience using the PoIA. Model 2 in Table 7 shows that the interaction term between the PoIA and integration experience is not significant, which suggests that the inertia caused by integration experience does not depend on the level of acquisition experience. Table 7. Logit predicting integration probability with the interaction between integration experience and acquisition experience (Model 1) or the PoIA and acquisition experience (Model 2) (1) (2) Variables Integration Integration Integration experience 0.923*** (0.223) Acquisition experience −0.049* 0.004 (0.029) (0.025) Integration experience × acquisition experience 0.002 (0.015) PoIA 3.178*** (0.768) PoIA × acquisition experience 0.033 (0.061) Relatedness 2.135*** 2.090*** (0.358) (0.354) Parallel acquisitions 0.369*** 0.398*** (0.083) (0.085) Target quality −0.356* −0.487*** (0.185) (0.180) Relative acquisition size 0.020 −0.023 (0.072) (0.075) Constant −1.555*** −2.799*** (0.423) (0.545) Observations 384 384 Pseudo R-squared 0.3980 0.4081 Log likelihood −130.647 −128.460 (1) (2) Variables Integration Integration Integration experience 0.923*** (0.223) Acquisition experience −0.049* 0.004 (0.029) (0.025) Integration experience × acquisition experience 0.002 (0.015) PoIA 3.178*** (0.768) PoIA × acquisition experience 0.033 (0.061) Relatedness 2.135*** 2.090*** (0.358) (0.354) Parallel acquisitions 0.369*** 0.398*** (0.083) (0.085) Target quality −0.356* −0.487*** (0.185) (0.180) Relative acquisition size 0.020 −0.023 (0.072) (0.075) Constant −1.555*** −2.799*** (0.423) (0.545) Observations 384 384 Pseudo R-squared 0.3980 0.4081 Log likelihood −130.647 −128.460 Note: Robust standard errors in parentheses. * P < 0.1; ** P < 0.05; *** P < 0.01. Table 7. Logit predicting integration probability with the interaction between integration experience and acquisition experience (Model 1) or the PoIA and acquisition experience (Model 2) (1) (2) Variables Integration Integration Integration experience 0.923*** (0.223) Acquisition experience −0.049* 0.004 (0.029) (0.025) Integration experience × acquisition experience 0.002 (0.015) PoIA 3.178*** (0.768) PoIA × acquisition experience 0.033 (0.061) Relatedness 2.135*** 2.090*** (0.358) (0.354) Parallel acquisitions 0.369*** 0.398*** (0.083) (0.085) Target quality −0.356* −0.487*** (0.185) (0.180) Relative acquisition size 0.020 −0.023 (0.072) (0.075) Constant −1.555*** −2.799*** (0.423) (0.545) Observations 384 384 Pseudo R-squared 0.3980 0.4081 Log likelihood −130.647 −128.460 (1) (2) Variables Integration Integration Integration experience 0.923*** (0.223) Acquisition experience −0.049* 0.004 (0.029) (0.025) Integration experience × acquisition experience 0.002 (0.015) PoIA 3.178*** (0.768) PoIA × acquisition experience 0.033 (0.061) Relatedness 2.135*** 2.090*** (0.358) (0.354) Parallel acquisitions 0.369*** 0.398*** (0.083) (0.085) Target quality −0.356* −0.487*** (0.185) (0.180) Relative acquisition size 0.020 −0.023 (0.072) (0.075) Constant −1.555*** −2.799*** (0.423) (0.545) Observations 384 384 Pseudo R-squared 0.3980 0.4081 Log likelihood −130.647 −128.460 Note: Robust standard errors in parentheses. * P < 0.1; ** P < 0.05; *** P < 0.01. 5. Discussion In this article, we have developed and tested a theory of how past decisional persistence and knowledge codification affect persistence in decision-making. We considered two fundamental factors related to learning—integration experience and knowledge codification—and we theorized and showed that whereas integration experience tends to generate persistence in decision-making, codification of knowledge of the integration process amplifies this persistence at low levels and then overcomes it at high levels. Moreover, we provided evidence that implies that persistence in decision-making is likely not related to learning—i.e., superior capabilities—but rather to inertia—i.e., persistence with a decision despite its negative performance consequences. This study offers several contributions. First, we help explain why a growing body of research on the performance implications of experience in strategic contexts offers mixed results (Barkema and Schijven, 2008). By analyzing how experience and knowledge codification interact, we avoid the common tendency to infer about the impact of experience simply by looking at its direct impact on the nature of subsequent decisions or performance. Instead, we propose and show that the impact of experience on decision-making persistence is contingent on other factors, such as knowledge codification; even more importantly, this moderating effect is nonlinear. This finding suggests that future research should look further at the contingency factors of experience to shed new light on the mixed results of experience on subsequent decisions and performance. Second, through our investigation, we respond to calls for a stronger focus on organizational theory within decision-making research. Gavetti and Rivkin (2007) argue that “the centrality of both organizations and decision making has been, to a great extent, lost” and that “work in organizational learning, despite having clear roots in the Carnegie School, has shifted away from issues of decision making” (p. 524). By studying the impact of experience and knowledge codification on the persistence of decisions, we retrain the spotlight on the relationship between decision-making and two fundamental factors of organizational learning. Third, in analyzing the effects of knowledge codification at different levels, we show that knowledge codification cannot be characterized solely as a source of inertia (i.e., coercive factor) or as a source of learning (i.e., enabling factor). Instead, knowledge codification likely changes from a magnifier of inertia to a source of learning as it increases. These results thus challenge the conventional wisdom that “portray[s] codification as a panacea, largely disregarding the inertia it entails, as stressed by a long line of prior work” (Heimeriks et al., 2012: 720). Rather, we find that codification simultaneously has pros and cons and that its inertial effects are overcome after a certain level. At the extreme, organizations without any knowledge are thus less likely to fall into decision inertia traps than those with a little knowledge codification. However, our findings also show that significant investments in knowledge overcome these inertial effects. Finally, our article is also germane to the literature that frames the effect of organizational acquisition experience on acquisition performance from a behavioral learning perspective (Haleblian and Finkelstein, 1999; Finkelstein and Haleblian, 2002). These studies have documented a U-shaped relationship between the number of acquisitions a firm has conducted and a focal acquisition’s performance. The theory behind this result hinges on a “negative transfer” process. After making their first acquisitions, inexperienced acquirers inappropriately generalize what they have learned to subsequent acquisitions, which are different from those previously conducted, thus hurting acquisition performance. More-experienced acquirers, however, learn to appropriately discriminate between their acquisitions and are less likely to unduly transfer routines and practices established in prior situations to new, dissimilar situations. Although this theory is appealing and has received empirical confirmation, it has investigated mostly the (indirect) link between experience and performance. In this study, we attempt to dig deeper into this process and study the effects of experiential and deliberate learning on the organizational decision process and how they affect performance in turn. We obtain results consistent with this theory. Thus, our results provide additional evidence consistent with the idea that firms tend to persist in using and applying old schemes and routines when making decisions related to organizational integration in the post-M&A phase. Nonetheless, complementing prior studies, we show that knowledge codification is the actual mechanism behind firms’ ability to recognize two different deals as different, and hence, it helps firms avoid the “negative transfer” effect caused by the accumulation of experience with a particular decision. Our results also have important managerial implications. First, acquirers that invest in knowledge codification can learn to overcome persistence in post-acquisition decisions. Therefore, managers should realize that increasing their firms’ knowledge levels decreases automatic triggers of predefined responses, reduces inertia, and encourages more-varied, effective decision-making processes. In addition, in the post-acquisition phase, managerial decisions need to be tailored to the focal acquisition. Moreover, the finding that the benefits of knowledge codification outweigh its costs only when a firm has sufficiently invested in knowledge codification is particularly important for acquiring firms. Acquirers vary significantly in the extent to which they develop specific tools based on their post-acquisition integration activities: some acquirers codify most activities in manuals; others codify few of them. An acquiring firm that has developed only a due diligence checklist, for example, differs greatly from one that has developed multiple manuals to describe the different activities in the post-merger phase, such as the conversion of information systems, the affiliation of human resources, financial evaluation spreadsheets, product training programs, and project management packages. Our findings suggest that investing in knowledge codification is worthwhile only if firms span a significant number of post-acquisition integration activities. However, such extensive codification is challenging for many acquirers, such as small firms with few resources to invest in developing manuals or those that acquire only a few targets throughout their history. Therefore, it would be interesting to investigate whether the difficulty they face in developing complete codifications of different M&A activities might be overcome, at least partially, by involving investment bankers, consulting firms, or strategic advisors. Our findings, however, suggest that it is not beneficial to begin a codification process superficially and merely following managerial fads. Our study has a number of limitations, some of which refer to the validity and generalizability of our results. In this single-industry study, we focus only on US bank mergers. As usual, the applicability of the findings to other samples in the same industry, other industries, or other geographic and institutional contexts should therefore be considered carefully. Other limitations arise from the need to triangulate and validate our results across different post-acquisition decisions. Additionally, acquisitions represent a particularly rare and complex type of strategic decision. Future studies should analyze whether similar results arise for strategic decisions that are less rare or complex. Most importantly, our data structure does not permit us to unambiguously establish causality, only conditional correlations. Finally, because of a lack of data, we do not study the role of knowledge articulation (Zollo and Winter, 2002). Studying whether knowledge articulation strengthens or weakens the persistence caused by integration experience represents an important avenue for future research. Nevertheless, our results can guide scholars in promising directions, namely, toward a better understanding of the antecedents and evolution of organizational decisions. In addition, we hope further research can provide new insights on the causes of persistence in decision-making and the learning mechanisms that might mitigate these forces. Researchers might also shed new light on the micro-processes that explain why knowledge codification generates inertia below a certain threshold and learning above it. Acknowledgments We are grateful to editor Fredrik Tell and two anonymous reviewers for their insightful comments and guidance. We also would like to thank Stefano Brusoni, Raffaele Conti, Phanish Puranam, as well as seminar participants at the SEI Faculty Workshop and the AOM Meeting for valuable comments on previous versions of the manuscript. Francesco Castellaneta acknowledges funding from Catolica Lisbon, and Giovanni Valentini from the “Agencia Estatal de Investigación” (AEI) of the Spanish Ministry of Economy and Competitiveness (Ref. ECO2015-ECO2015-71173-P) (AEI/FEDER, UE). Footnotes 1 The data set we use in this article has been used in two other published papers: (Zollo, 2009) and (Zollo and Singh, 2004). However, the research questions pursued in these two papers clearly differ from the one addressed in this article. The two papers explain M&A performance (i.e., M&A performance is the dependent variable), while in this article, we explain the post-acquisition integration decision (i.e., the integration decision is the dependent variable) instead. 2 In robustness results (available upon request), we do not dichotomize the variable but use a dependent variable that takes the following values: 0 if “few or no features were aligned or centralized”; 1 “if only selected systems, procedures, or products were aligned or centralized”; 2 if “many but not all systems, procedures, and products were aligned or centralized”; and 3 if “all systems, procedures, and products were completely integrated.” When using an ordered probit model, we obtain similar results. 3 To avoid zero values for the argument of the log, we follow Helsel (2005) and add the smallest quantity that makes the lognormal probability plot approximately linear. 4 Results (available upon request) are robust to the inclusion of year dummies. 5 Due to missing values for ROA, we include in the analysis only 26 acquiring banks that completed 259 acquisitions overall. 6 This result is indirectly clear from previous regressions but confirmed when directly regressing Deviation on integration experience (results available from the authors). References Adler P. S. , Borys B. ( 1996 ), ‘ Two types of bureaucracy: enabling and coercive ,’ Administrative Science Quarterly , 41 ( 1 ), 61 – 89 . Google Scholar CrossRef Search ADS Amburgey T. , Miner A. ( 1992 ), ‘ Strategic momentum: the effects of repetitive, positional, and contextual momentum on merger activity ,’ Strategic Management Journal , 13 ( 5 ), 335 – 348 . Google Scholar CrossRef Search ADS Amburgey T. L. , Kelly D. , Barnett W. P. ( 1993 ), ‘ Resetting the clock: the dynamics of organizational change and failure ,’ Administrative Science Quarterly , 38 ( 1 ), 51 – 73 . Google Scholar CrossRef Search ADS Balconi M. , Pozzali A. , Viale R. ( 2007 ), ‘ The ‘codification debate’ revisited: a conceptual framework to analyze the role of tacit knowledge in economics ,’ Industrial and Corporate Change , 16 ( 5 ), 823 – 849 . Google Scholar CrossRef Search ADS Barkema H. G. , Schijven M. ( 2008 ), ‘ How do firms learn to make acquisitions? A review of past research and an agenda for the future ,’ Journal of Management , 34 ( 3 ), 594 – 634 . Google Scholar CrossRef Search ADS Baum J. A. , Li S. X. , Usher J. M. ( 2000 ), ‘ Making the next move: how experiential and vicarious learning shape the locations of chains’ acquisitions ,’ Administrative Science Quarterly , 45 ( 4 ), 766 – 801 . Google Scholar CrossRef Search ADS Becker M. C. ( 2004 ), ‘ Organizational routines: a review of the literature ,’ Industrial and Corporate Change , 13 ( 4 ), 643 – 678 . Google Scholar CrossRef Search ADS Becker M. C. , Lazaric N. , Nelson R. , Winter S. ( 2005 ), ‘ Applying organizational routines in understanding organizational change ,’ Industrial and Corporate Change , 14 ( 5 ), 775 – 791 . Google Scholar CrossRef Search ADS Benner M. J. , Tushman M. ( 2002 ), ‘ Process management and technological innovation: a longitudinal study of the photography and paint industries ,’ Administrative Science Quarterly , 47 ( 4 ), 676 – 707 . Google Scholar CrossRef Search ADS Bergh D. D. , Lim E. N. ( 2008 ), ‘ Learning how to restructure: absorptive capacity and improvisational views of restructuring actions and performance ,’ Strategic Management Journal , 29 ( 6 ), 593 – 616 . Google Scholar CrossRef Search ADS Castellaneta F. , Salvato C. ( forthcoming ), ‘ Culminating events and time working together in top management teams: insights from private equity ,’ Long Range Planning . Castellaneta F. , Zollo M. ( 2015 ), ‘ The dimensions of experiential learning in the management of activity load ,’ Organization Science , 26 ( 1 ), 140 – 157 . Google Scholar CrossRef Search ADS Cowan R. , David P. A. , Foray D. ( 2000 ), ‘ The explicit economics of knowledge codification and tacitness ,’ Industrial and Corporate Change , 9 ( 2 ), 211 – 253 . Google Scholar CrossRef Search ADS Cowan R. , Foray D. ( 1997 ), ‘ The economics of codification and the diffusion of knowledge ,’ Industrial and Corporate Change , 6 ( 3 ), 595 – 622 . Google Scholar CrossRef Search ADS Cyert R. M. , March J. G. ( 1963 ), A Behavioral Theory of the Firm . Prentice Hall : New York, NY . Datta D. K. ( 1991 ), ‘ Organizational fit and acquisition performance: effects of post-acquisition integration ,’ Strategic Management Journal , 12 ( 4 ), 281 – 297 . Google Scholar CrossRef Search ADS DeYoung R. , Evanoff D. D. , Molyneux P. ( 2009 ), ‘ Mergers and acquisitions of financial institutions: a review of the post-2000 literature ,’ Journal of Financial Services Research , 36 ( 2–3 ), 87 – 110 . Google Scholar CrossRef Search ADS Dimov D. , de Holan P. M. , Milanov H. ( 2012 ), ‘ Learning patterns in venture capital investing in new industries ,’ Industrial and Corporate Change , 21 ( 6 ), 1389 – 1426 . Google Scholar CrossRef Search ADS Finkelstein S. , Haleblian J. ( 2002 ), ‘ Understanding acquisition performance: the role of transfer effects ,’ Organization Science , 13 ( 1 ), 36 – 47 . Google Scholar CrossRef Search ADS Foray D. , Steinmueller W. E. ( 2003 ), ‘ The economics of knowledge reproduction by inscription ,’ Industrial and Corporate Change , 12 ( 2 ), 299 – 319 . Google Scholar CrossRef Search ADS Gavetti G. , Levinthal D. A. ( 2000 ), ‘ Looking forward and looking backward: cognitive and experiential search ,’ Administrative Science Quarterly , 45 ( 1 ), 113 – 137 . Google Scholar CrossRef Search ADS Gavetti G. , Rivkin J. W. ( 2007 ), ‘ On the origin of strategy: action and cognition over time ,’ Organization Science , 18 ( 3 ), 420 – 439 . Google Scholar CrossRef Search ADS Groves R. M. , Cialdini R. B. , Couper M. P. ( 1992 ), ‘ Understanding the decision to participate in a survey ,’ Public Opinion Quarterly , 56 ( 4 ), 475 – 495 . Google Scholar CrossRef Search ADS Gulati R. ( 1995 ), ‘ Social structure and alliance formation patterns: a longitudinal analysis ,’ Administrative Science Quarterly , 40 ( 4 ), 619 – 652 . Google Scholar CrossRef Search ADS Gulati R. , Gargiulo M. ( 1999 ), ‘ Where do interorganizational networks come from? ,’ American Journal of Sociology , 104 ( 5 ), 1439 – 1493 . Google Scholar CrossRef Search ADS Haans R. F. , Pieters C. , He Z. ( 2016 ), ‘ Thinking about U: theorizing and testing U and inverted U‐shaped relationships in strategy research ,’ Strategic Management Journal , 37 ( 7 ), 1177 – 1195 . Google Scholar CrossRef Search ADS Hagendorff J. , Keasey K. ( 2009 ), ‘ Post-merger strategy and performance: evidence from the US and European banking industries ,’ Accounting and Finance , 49 ( 4 ), 725 – 751 . Google Scholar CrossRef Search ADS Håkanson L. ( 2007 ), ‘ Creating knowledge: the power and logic of articulation ,’ Industrial and Corporate Change , 16 ( 1 ), 51 – 88 . Google Scholar CrossRef Search ADS Haleblian J. , Finkelstein S. ( 1999 ), ‘ The influence of organizational acquisition experience on acquisition performance: a behavioral learning perspective ,’ Administrative Science Quarterly , 44 ( 1 ), 29 – 56 . Google Scholar CrossRef Search ADS Haleblian J. , Kim J. J. , Rajagopalan N. ( 2006 ), ‘ The influence of acquisition experience and performance on acquisition behavior: evidence from the U.S. Commercial banking industry ,’ Academy of Management Journal , 49 ( 2 ), 357 – 370 . Google Scholar CrossRef Search ADS Hannan T. H. , Pilloff S. J. ( 2009 ), ‘ Acquisition targets and motives in the banking industry ,’ Journal of Money, Credit and Banking , 41 ( 6 ), 1167 – 1187 . Google Scholar CrossRef Search ADS Heimeriks K. H. , Schijven M. , Gates S. ( 2012 ), ‘ Manifestations of higher-order routines: the underlying mechanisms of deliberate learning in the context of postacquisition integration ,’ Academy of Management Journal , 55 ( 3 ), 703 – 726 . Google Scholar CrossRef Search ADS Helsel D. R. ( 2005 ), Nondetects and Data Analysis . John Wiley and Sons : New York, NY . Horrace W. C. , Oaxaca R. L. ( 2006 ), ‘ Results on the bias and inconsistency of ordinary least squares for the linear probability model ,’ Economics Letters , 90 ( 3 ), 321 – 327 . Google Scholar CrossRef Search ADS Hoskisson R. E. , Hitt M. A. ( 1988 ), ‘ Strategic control systems and relative R&D investment in large multiproduct firms ,’ Strategic Management Journal , 9 ( 1 ), 605 – 621 . Google Scholar CrossRef Search ADS Kale P. , Singh H. ( 2007 ), ‘ Building firm capabilities through learning: the role of the alliance learning process in alliance capability and firm-level alliance success ,’ Strategic Management Journal , 28 ( 10 ), 981 – 1000 . Google Scholar CrossRef Search ADS Katz M. , Shapiro C. ( 1985 ), ‘ Network externalities, competition and compatibility ,’ American Economic Review , 75 ( 3 ), 424 – 440 . Levinthal D. A. , March J. G. ( 1993 ), ‘ The myopia of learning ,’ Strategic Management Journal , 14 ( S2 ), 95 – 112 . Google Scholar CrossRef Search ADS Levitt B. , March J. G. ( 1988 ), ‘ Organizational learning ,’ Annual Review of Sociology , 14 ( 1 ), 319 – 340 . Google Scholar CrossRef Search ADS March J. , Simon H. A. ( 1958 ), Organizations . Wiley : New York, NY . Nelson R. R. , Winter S. G. ( 1982 ), An Evolutionary Theory of Economic Change . Belknap : Cambridge, MA . Nonaka I. ( 1994 ), ‘ A dynamic theory of organizational knowledge creation ,’ Organization Science , 5 ( 1 ), 14 – 37 . Google Scholar CrossRef Search ADS Pentland B. T. , Feldman M. S. ( 2005 ), ‘ Organizational routines as a unit of analysis ,’ Industrial and Corporate Change , 14 ( 5 ), 793 – 815 . Google Scholar CrossRef Search ADS Prencipe A. , Tell F. ( 2001 ), ‘ Inter-project learning: processes and outcomes of knowledge codification in project-based firms ,’ Research Policy , 30 ( 9 ), 1373 – 1394 . Google Scholar CrossRef Search ADS Romme G. L. , Zollo M. , Berends P. ( 2010 ), ‘ Dynamic capabilities, deliberate learning and environmental dynamism: a simulation model ,’ Industrial and Corporate Change , 19 ( 4 ), 1271 – 1299 . Google Scholar CrossRef Search ADS Schulz M. ( 1998 ), ‘ Limits to bureaucratic growth: the density dependence of organizational rule births ,’ Administrative Science Quarterly , 43 ( 4 ), 845 – 876 . Google Scholar CrossRef Search ADS Shaver J. M. , Mitchell W. , Yeung B. ( 1997 ), ‘ The effect of own-firm and other-firm experience on foreign direct investment survival in the United States, 1987–1992 ,’ Strategic Management Journal , 18 ( 10 ), 811 – 824 . Google Scholar CrossRef Search ADS Snell S. A. ( 1992 ), ‘ Control theory in strategic human resource management: the mediating effect of administrative information ,’ Academy of Management Journal , 35 ( 2 ), 292 – 327 . Sydow J. , Schreyögg G. , Koch J. ( 2009 ), ‘ Organizational path dependence: opening the black box ,’ Academy of Management Review , 34 ( 4 ), 689 – 709 . Szulanski G. ( 1996 ), ‘ Exploring internal stickiness: impediments to the transfer of best practice within the firm ,’ Strategic Management Journal , 17 ( 1 ), 27 – 43 . Google Scholar CrossRef Search ADS Zander U. , Kogut B. ( 1995 ), ‘ Knowledge and the speed of the transfer and imitation of organizational capabilities: an empirical test ,’ Organization Science , 6 ( 1 ), 76 – 92 . Google Scholar CrossRef Search ADS Zollo M. ( 2009 ), ‘ Superstitious learning with rare strategic decisions: theory and evidence from corporate acquisitions ,’ Organization Science , 20 ( 5 ), 894 – 908 . Google Scholar CrossRef Search ADS Zollo M. , Singh H. ( 2004 ), ‘ Deliberate learning in corporate acquisitions: post-acquisition strategies and integration capability in U.S. bank mergers ,’ Strategic Management Journal , 25 ( 13 ), 1233 – 1256 . Google Scholar CrossRef Search ADS Zollo M. , Winter S. G. ( 2002 ), ‘ Deliberate learning and the evolution of dynamic capabilities ,’ Organization Science , 13 ( 3 ), 339 – 351 . Google Scholar CrossRef Search ADS © The Author 2017. Published by Oxford University Press on behalf of Associazione ICC. All rights reserved. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices)

Journal

Industrial and Corporate ChangeOxford University Press

Published: Oct 30, 2017

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off