Abstract
This study investigates what leads managers to allocate constrained cognitive effort toward new versus familiar aspects of a business. Specifically, we explore advice giving by venture capital firms (VCs) to their portfolio companies, distinguishing between business topics on which a VC has advised other ventures in the past and topics new to the VC that may be outside its areas of expertise. We use both demand-side (venture-driven) and supply-side (VC-driven) perspectives to offer a novel theory about the antecedents of cognitive effort underlying advice giving. Empirical tests in a unique data set of French VCs show that both perspectives explain important aspects of advice-giving dynamics for VCs. VCs facing dynamic environments and capacity constraints strongly respond to stimuli from ventures, but VCs also adjust their behavior as they accumulate experience in ways that reflect both expanding confidence in their ability to add value and concerns about overextension of their efforts, depending on the valence of VC experience. Our findings provide important insights to the antecedents of cognitive effort and to research on the VC–venture relationship by exploring the dynamics of how advice-giving relationships evolve over time as VCs gain experience.
Keywords
Managerial attention is one of the most constrained and important resources within organizations (Penrose, 1959; Simon, 1945). Therefore, deciding where to allocate cognitive effort—including participation in decision making, offering advice, and dedicating attention—is one of the paramount issues facing an organization (Adner & Helfat, 2003; Teece, 2012). Lack of attention to important stimuli or misdirected attention can result in organizational decline, failure to respond to important threats, and even catastrophes (Corwin & Coughenour, 2008; Joseph & Ocasio, 2012). The importance of directing cognitive effort is well established in the literature. What influences the direction of cognitive effort, however, is much less established. The limited existing work exploring antecedents of the direction of cognitive effort focus either on massive environmental changes that affect all firms (e.g., Barr, Stimpert, & Huff, 1992) or on stable, microlevel characteristics of the manager that affect cognition (e.g., Gerstner, König, Enders, & Hambrick, 2013). In effect, there is a broad lacuna in our understanding of the antecedents of cognitive effort, particularly in dynamic environments where managers must shift their attention between new and familiar aspects of the business. Research highlights important adaptive implications of attention to developing new technologies (Eggers & Kaplan, 2009), advising in new business models (Cho & Hambrick, 2006; Gerasymenko, De Clercq, & Sapienza, 2015), and responding to new representations of the environment (Barr et al., 1992; Nadkarni & Barr, 2008). While cognitive effort directed toward new business problems is certainly important, managers and board members face important constraints on their time and attention (Penrose, 1959; Ocasio, 1997) and must balance their efforts between new and familiar issues. Understanding why decision makers shift their cognitive effort and what impacts these choices will provide greater insight into the heterogeneous behavior of firms and can lead to new insights into firm-level evolution.
We seek to explore the dynamics of cognitive effort by incorporating contextual factors that affect the behavior of specific decision makers, such as boards of directors, following recent calls by prominent scholars (e.g., Argote & Greve, 2007: 344). We build and test our theory about the factors that affect changes in advice giving—a specific manifestation of cognitive effort—in the context of venture capital firms (VCs) providing advice to their venture investments. Specifically, we explore the question of when and why a VC partner would provide advice to a given venture in either a domain familiar to the VC or a comparatively new domain. While we do not equate advice giving with cognition, we see cognitive effort as an important precursor to advice giving. For instance, advice giving by a VC involves cognitive processing, interpretation, and transfer of information to the managers of a given venture. Analyzing VC advising may, therefore, provide an analogous but distinct window into the underlying construct of cognition in much the same way as other commonly used proxies, such as letters to shareholders (Kaplan, 2008; Nadkarni & Barr, 2008) and public speeches (Chatterjee & Hambrick, 2007).
Providing advice and insight to their venture investments is one of the central tasks for VCs, along with fundraising, investing, and monitoring ventures’ performance (De Clercq, Fried, Lehtonen, & Sapienza, 2006). Given that VCs typically invest in highly uncertain early-stage ventures operating in dynamic environments and receive feedback from current venture activities as well as from their experiences (De Clercq & Sapienza, 2005), we explore two broad types of explanations that are especially appropriate in the context of the VC–venture relationship. First, we look at how the direction of advice giving is affected by the current challenges and signals offered to the VC by their ventures. This builds on the idea that new-venture dynamics are incredibly fast-paced (Sapienza, 1992), meaning that advisors may have neither time nor incentive to develop organizational routines to respond to the challenges in front of them (Arthurs & Busenitz, 2006). Second, we explore the extent to which VCs change their advice-giving behavior as they accumulate experience and the environment around them changes (Levinthal & March, 1993; Posen & Levinthal, 2012). Our approach expands on the existing limited research on learning by VCs that has focused largely on how VCs learn to select better ventures based on experience (Shepherd, Zacharakis, & Baron, 2003) and provides a more holistic understanding of the dynamics of advice giving.
We test our theory using novel data on the advice giving of French VC partners. Our results show two primary levels of findings. First, we show that proximate feedback on current ventures strongly affects advice-giving behavior. Interestingly, challenges for the current venture are positively related to increased advice giving in new domains, as would be predicted by typical responses to problems, but venture challenges also increase advising in familiar domains. This suggests that problem-driven search leads to increases in both familiar and novel effort in order to find solutions. Meanwhile, physical proximity leads VC partners to intensify their advice giving in familiar domains (presumably because all ventures need help and the cost of offering advice is lower) but has no effect on advising in new domains as there is no clear stimulus to trigger search. Second, we also find important effects for the accumulation of experience by the VC that are consistent with adaptive behavior through learning. Specifically, we find that VCs tend to give less advice in both new and familiar domains as they accumulate experience. This effect is especially true for unsuccessful experience (as VCs retreat toward core areas of expertise where they feel more confident in their knowledge) and much less true for successful experience, where VCs may feel emboldened to increase involvement. Together, these perspectives offer specific insights into how the direction of advice giving toward new versus familiar domains is affected both by proximate stimuli in a reactive manner and through the accumulation of experience in a more proactive manner.
This study makes multiple important contributions to the literature. First, this study advances our understanding of the advising role of boards of directors and provides novel insights into the evolving factors that stimulate and restrain directors’ advising effort. Importantly, our findings reveal that both firm-centric and director-centric factors influence the amount and direction of advice, suggesting that both should be considered concurrently. Second, this study joins the limited work on the antecedents of managerial cognition (Cho & Hambrick, 2006; Gerstner et al., 2013; Nadkarni & Barr, 2008). Our results suggest that exploitative (familiar domains) and explorative (new domains) efforts are not necessarily two ends of one continuum but are two distinct sets of activities that an organization can pursue simultaneously. Finally, this study makes important contributions to the literature on the VC–venture relationship. This study illustrates why not all VC advice may carry the same value and thereby provides important perspective to entrepreneurs when considering a VC’s advice. Specifically, the study underscores that entrepreneurs should be mindful of potential cognitive biases generated through VC experience as well as account for potential limitations to effective advising if they choose to raise funding from a foreign VC.
Theory and Hypotheses
Before articulating specific hypotheses about how advice-giving behavior relates to the dynamics of venture challenges and the evolution of VC experience, we seek to accomplish two major tasks in this section. First, we clarify the link between advice giving and cognition. Second, we seek to understand and synthesize what we know about advice giving in the VC–venture relationship.
Advice Giving and Cognitive Effort
Management research has long been interested in cognition, but the challenges to directly measuring cognition are significant. As a result, scholars typically rely on proxies that carry a strong cognitive component but do not directly equate to cognition. Most prominent in recent years has been the use of statements in letters to shareholders (e.g., Kaplan, 2008; Nadkarni & Barr, 2008) but also includes the use of earnings call transcripts (e.g., Mayew & Venkatachalam, 2012) and public speeches (e.g., Chatterjee & Hambrick, 2007). None is a perfect proxy, but we argue that advice giving provides an analogous window into cognition because it requires cognitive information processing, interpretation, and knowledge-transfer effort.
For example, research shows that a director’s advising was more effective if he or she could draw on analogous direct experience when overseeing a different firm (McDonald, Westphal, & Graebner, 2008) and if the firm operated in a stable environment where accumulated knowledge would still be helpful (Carpenter & Westphal, 2001). Such boundary conditions emerge because managers may rely on proven heuristics (Eisenhardt, Furr, & Bingham, 2010) that facilitate improved cognitive processing (Lavie & Rosenkopf, 2006). The recognition of advice giving as a core cognitive task for managers means that our approach to understanding the focus of managerial advice giving on novel versus familiar aspects of the business contributes to the limited literature studying the antecedents of cognition (Argote & Greve, 2007; Gavetti, Greve, Levinthal, & Ocasio, 2012).
VC Advice Giving in Novel and Familiar Domains
VCs are institutional and professional investors who take an equity stake in exchange for their financial investment and are actively involved in investee firms (Amit, Brander, & Zott, 1998; Sahlman, 1990). VCs are very different from banks and typical financial investors that provide only financing and monitoring. VCs are specialized investors that create value by selecting promising ventures (often early-stage firms in high-tech sectors) and by providing extensive support following the investment. This support takes the form of network connections and enhanced reputation (Hsu, 2004) as well as the knowledge and capabilities of the VC partners (Brander, Amit, & Antweiler, 2002; Sapienza, 1992). VCs typically offer intensive and continuous advice on a variety of topics, including strategy, pricing, fund-raising, and networking (MacMillan, Kulow, & Khoylian, 1989). This added value from VCs helps VC-backed ventures have more successful commercialization strategies (Hsu, 2006) and a shorter time to market for their technology (Hellmann & Puri, 2002). Thus, the added value transferred in the VC–venture relationship is a key element to understanding venture success.
Importantly, research highlights that the value of a VC’s efforts may differ across domains (Busenitz, Fiet, & Moesel, 2004) and that the added value is constrained by the bounded attention of VC partners. For instance, Jackson, Bates, and Bradford (2012) demonstrate that VC advising significantly improves venture performance and the VC’s internal rate of return (IRR), but VC advice giving becomes spread too thin past a portfolio-size threshold, leading to IRR decline. As a result, VCs typically specialize in a limited set of domains, both for investments (Sapienza & Manigart, 1996) and advising (Busenitz et al., 2004). This quote from one of our semistructured interviews with VC partners both highlights this degree of VC knowledge specialization and begins to offer insight into the important question of when and why a VC may choose to focus its advice-giving effort on relatively new domains:
While our general partners are pretty comfortable with technological aspects of ventures we invest in, we tend to take a more hands-off approach when it comes to managing human capital and leadership roles within ventures’ teams. In most cases we rely on co-investors who take a more proactive role in changing management team composition and hiring when it was necessary; yet, we had to intervene into managerial issues in one of our recent investments because we did not have a co-investor who could do it. We did not feel as comfortable in this role as we are when it comes to discussing technological implications and marketing issues with our portfolio companies, but we could not leave the founder of that venture alone because he clearly needed help and support.
This quote underscores potential drivers of advice-giving dynamics that are central to our theory, such as problems of VCs’ portfolio ventures and VC experience. Our study seeks to expand upon this dynamic by exploring when VCs provide advice in new domains and when they focus on more familiar domains.
For VC partners used to drawing on their own specialized knowledge and experience, providing advice in novel domains implies investing time and resources in acquiring new knowledge. Advice giving in novel domains is likely associated with greater uncertainty around the advice’s effectiveness, in large part because of the complexity of the issues in most domains. Our interviews suggested that advising in finance would typically encompass explanations of financial reporting systems, revision of key performance indicators, discussion of expenditures to reach milestones, and financial planning. Advising about business models typically involved representing the venture’s business model, researching business models used by competitors, explicating the venture’s value-add proposition, and building a roadmap for adjusting the existing business model (e.g., resources, leading change). Thus, adding real value in any given domain requires knowledge and understanding of the intricacies of each domain. While offering advice in familiar domains is likely to require cognitive effort for VCs, advising in new domains can be expected to require even more cognitive effort.
Venture-Driven Determinants of Advice Giving
VCs aim to add value in domains where ventures lack expertise (Colombo & Grilli, 2010), helping founders solve new and challenging problems, especially during the early developmental stages of ventures (Rosenstein, Bruno, Bygrave, & Taylor, 1993) where task uncertainty is high (Sapienza & Gupta, 1994: 1622). In addition, as advising typically builds on the VC’s specific local knowledge and network, the location of portfolio companies (PFCs) likely influences the extent and the domains of VC advising. This suggests that VCs may make critical advice-giving decisions based on the needs and locations of their ventures. Such factors can be thought of as demand-side factors that alter the venture’s demand for VC advising. We explore these factors below.
Venture problems and advice giving
Managers respond to problems by searching for new solutions, which is why prior research has focused on R&D spending when studying problem-driven search (e.g., Greve, 2003). In other words, problems spur exploratory efforts to find solutions (Greve, 2003). However, Cyert and March (1963) explicitly state that the organizational response to a problem is to engage in “simple-minded” search, where the organization presumes that the “new solution can be found ‘near’ an old one” (p. 170). Thus, firms may begin responding to problems by looking for local solutions, turning to increasingly distant options only when initial efforts do not solve the problem. This suggests that problems may increase managerial effort and that this effort may be directed at both new and familiar domains as managers seek suitable solutions.
Translated to the VC–venture context, problems in a given venture should lead the VC partner to increase advice giving in both familiar and novel domains. For a new venture, a substantial business model change serves as strong evidence of serious problems. This happens when the original business model fails to deliver its expected value (Nicholls-Nixon, Cooper, & Woo, 2000), which leads ventures to substantially change business models (Zott & Amit, 2007). While VCs could proactively advise ventures to adjust their business model before a problem fully manifests, we argue that a VC would not push for a business model change without having recognized a problem that limits the venture’s long-term potential. It is this underlying problem—measured through the proxy of business model change—that we argue is driving problem-driven search. Moreover, our interviews suggest that VCs frequently react to problems instead of foreseeing them ahead of time. As some VC partners explained, they faced less resistance from ventures’ CEOs when suggesting changes after evidence of performance deterioration was present. Given that making a substantial business model change involves changes in different domains, we expect that the problems precipitating business model change within a venture will lead the VC to deepen advice giving in familiar domains (local search). In addition, because implementing such changes requires coming up with creative combinations and solutions, it is likely to spur advice giving in novel domains (distant search) as the VC and venture jointly try to find a new business model. 1
Hypothesis 1a: A substantial change in the venture’s business model will be positively related to the amount of advice that the VC partner subsequently offers in new domains.
Hypothesis 1b: A substantial change in the venture’s business model will be positively related to the amount of advice that the VC partner subsequently offers in familiar domains.
Proximity and advice giving
Because advice giving relies to a great extent on in-person VC–venture interaction, physical proximity between the VC and the venture is of paramount importance. That is because VCs need to closely monitor ventures and frequently interact with entrepreneurs after investing in business (Cumming & Dai, 2010) in order to have in-depth understanding of different issues and to be able to offer valuable advice. More frequent interactions with ventures located in the home country enable VCs to identify competence gaps more efficiently and aim their advising to fill in those needs. At the same time, proximity also provides the opportunity for ventures to ask for more assistance. Substantial physical distance restricts the VC’s ability to do so (Lerner, 1995). VCs investing locally can leverage their local networks to provide effective advice for ventures.
By contrast, VCs investing in foreign ventures need to overcome cultural and institutional gaps (Hegde & Tumlinson, 2014). The cultural distance that captures national differences in cognitive patterns and values may create communication problems that hinder information sharing (Dai, Jo, & Kassicieh, 2012; Li, Vertinsky, & Li, 2014) and decrease trust. Institutional differences that are related to legal and other transaction costs may also be associated with incongruence in institutional practices and higher information asymmetry, thereby hindering the VC–venture advising relationship. All of these factors suggest that distance may constrain advice giving. We expect these challenges for VCs to be exacerbated for advice giving in novel domains: VCs may be less willing to offer advice in unfamiliar areas to foreign ventures due to constraints on their locally based network, and ventures may be less likely to demand such advice from VCs due to lack of trust. As a result, an increase in the number of local ventures in the VC’s portfolio is likely to lead to greater VC focus on advising in familiar domains. We therefore expect that a French-based VC will more actively advise in familiar domains as a number of PFCs located in France compared to foreign ventures increases.
Hypothesis 2: An increase in the proportion of ventures located in the same country as the VC will be positively associated with the amount of advice that the VC partner offers in familiar domains.
VC-Driven Determinants of Advice Giving
While VCs may strongly react to proximate, demand-side factors when determining how to allocate scarce advice-giving resources as discussed above, we seek to build on the limited literature studying learning by VCs to suggest that supply-side (VC) changes in experience and capability will have significant implications for the advice-giving behavior of VCs. As active investors, VCs’ experience is of paramount importance in the process of identifying, selecting, and advising ventures (De Clercq & Dimov, 2008). Although VC partners bring prior personal experience to the management team, their collective experience as a VC carries substantial weight in investment management process (Dimov & Martin de Holan, 2010). For these reasons, we focus on the volume and nature of VCs’ experience, and constraints on the VCs’ ability to deploy that experience, as supply-side factors that will affect advice-giving dynamics.
Experience and advice giving
As VCs accumulate experience, they engage in a learning process that helps them build both expertise and cognitive heuristics. This is in line with prior research on startups and other organizations (Bingham & Eisenhardt, 2011; Bingham & Haleblian, 2012) as well as with research in cognitive psychology (Kolb, 2014). This is also consistent with Shepherd et al.’s (2003) suggestion that increased VCs’ task experience tends to lead them to rely more on previously acquired knowledge and heuristics. Experience helps VC partners develop their personal expertise in whichever aspects of their prior ventures they engaged through a standard process of experience development.
Specifically, by having been involved in previous ventures through their entire lifecycle, VCs gain a more in-depth understanding of the different milestones and needs of ventures. Dealing with repeated decision situations leads to learning-curve effects (Argote, 2012) as problem definitions, alternative responses, and outcomes become encoded into knowledge that facilitates similar future decisions (De Clercq & Dimov, 2008). Based on its prior active experience, a VC can have a more nuanced understanding of a specific venture and thereby can focus advice on some of the most critical aspects that could affect the venture’s success. Greater previous exit experience provides VCs with opportunities to establish more accurate analogies between past exits and current investments, and thereby better discern the impact of advice in different domains on the success of their investments. Moreover, pressures stemming from path dependency and increased opportunity cost for the VC partner for providing advice in novel domains will likely lead VCs to increasingly focus their cognitive efforts on familiar domains as they gain experience.
Hypothesis 3: Higher volumes of previous exit experience by the VC will be positively associated with the amount of advice that the VC partner offers in familiar domains.
Based on their prior successes and failures, the VC firm managers may engage in group discussions and decision making, and make joint interpretations and draw collective inferences about the type of strategies that work in different situations and the effectiveness of different advice in interacting with PFC managers (Dimov & Martin de Holan, 2010). This implies that in deciding what advice to offer and in what domain, VC partners can rely on the experience and insights derived from different outcomes of previous investments. Yet, learning from successful and failed experience may evoke different learning.
VCs with a significant amount of positive previous exit experience—like managers in any other organization—may view this as validation of prior behavior, which should reinforce a similar behavior (Shepherd et al., 2003). By contrast, VCs with a preponderance of failed exit experience are more likely to view these prior failures as due (at least in part) to bad advice that they may have offered, making them more inclined to abandon advice-giving domains in which they had previously been active. For example, if a VC previously gave advice most often in strategy, marketing, and networking, but saw most of its prior ventures fail, the VC is likely to give up providing advice in whichever of those three domains it feels least qualified to offer advice. In terms of advice giving in familiar domains, therefore, the effect is relatively clear: Higher levels of success versus failure leads VCs to continue offering advice in familiar domains.
Hypothesis 4a: Higher levels of successful (versus unsuccessful) previous exit experience by the VC will be positively related to the amount of advice that the VC partner offers in familiar domains.
The effect of successful and failed exit experience on advice giving in new domains, however, is much less clear. On one hand, theories of managerial satisficing suggest that success discourages managers from making changes and leads to satisficing and complacency (Ellis, Mendel, & Nir, 2006). Failed experience is more likely to lead to a change in behavior in an effort to improve subsequent outcomes (Cyert & March, 1963). VCs may fall into a satisficing mode, preferring to rely on already acquired and tested knowledge in their interaction with ventures. Because successes instill greater confidence in one’s own knowledge (versus failures), VCs experiencing greater levels of success may be less inclined to search for new ways of adding value to their subsequent PFCs. Similar to other organizations, VCs are not exempt from self-attribution bias. Zacharakis, Meyer, and DeCastro (1999) found that VCs tend to attribute ventures’ success to internal (management) factors, whereas they put the blame on exogenous, out-of-control conditions for failures. Such self-attribution is likely to reinforce the VC’s confidence in its existing advice-giving expertise and reduce the need for exploration. Thus, successful exit experience may decrease the likelihood that a VC partner provides advice in novel domains.
A contrasting perspective on VC reaction to success (versus failure) is based largely on the implications of slack resources. While there is little research on the effect of slack resources from prior successes on VC-specific behavior, there is significant evidence from the broader literature. Success creates slack resources, which in turn encourages firms to explore new domains through slack search (Greve, 1998; Levinthal & March, 1993). Slack has been linked to exploratory efforts (Argote & Greve, 2007; Sharfman, Wolf, Chase, & Tansik, 1988), suggesting that slack provides a buffer versus the risk that diminishes the potential downside of such investments (Levinthal & March, 1993; Sidhu, Volberda, & Commandeur, 2004). In the case of VC advice giving, successful exit experience affects resources through increased profitability (Bayar & Chemmanur, 2011; Jain, Jayaraman, & Kini, 2008), which may create the resource buffer allowing VCs to explore new domains. Successful exit experience also has an indirect effect on attention to novel domains through VC confidence. Given that psychological research has shown that confidence can increase willingness to take risks (Kahneman & Lovallo, 1993; Krueger & Dickson, 1994), increased confidence from success may increase willingness to engage in novel activities. The idea that success may encourage exploration of new domains more than failure may be truer in the VC context than elsewhere, largely because VC partners expect low levels of success from their investments. Such “normalization” of failure may decrease the negative emotions associated with failures (Shepherd, Wiklund, & Haynie, 2009), which would lead VCs to react less to failures, meaning that they may actually change behavior more from successes. Our next hypothesis is therefore formulated as follows:
Hypothesis 4b: Higher levels of successful (versus unsuccessful) previous exit experience by the VC will be positively related to the amount of advice that the VC partner offers in novel domains.
Constraints to advice giving
Finally, cognition-related theories build on the idea that managers are boundedly rational, which includes being unable to be aware of and devote attention to every possible aspect of the business (Gavetti & Levinthal, 2000). As earlier research indicated, organizations may particularly experience constraints stemming from bounded rationality when managers experience cognitive or attention overload, limiting the ability of managers to effectively extend additional cognitive effort. We consider the effect that such load may have on effort toward new and familiar domains. A behavioral perspective would suggest that when cognitive constraints are reduced as a result of lower cognitive load, the opportunity for decision maker effort and advice giving should expand. Given that VC partners likely believe that their advice-giving activities improve outcomes for their ventures (why else would they engage in such activity?), we expect that a decrease in constraints should result in increased VC partner effort. In the absence of feedback on whether there are specific problems with a venture or changes in slack as discussed above, it makes sense that such increased advice-giving effort would be focused on domains where the VC partner already has specialized experience. In particular, VC firms face a significant trade-off in terms of VC size and span of control.
On the one hand, given the high levels of uncertainty surrounding early-stage ventures, VCs have an incentive to invest in as many ventures as possible to increase the chances that at least one venture is successful. On the other hand, given the relationship between higher levels of active VC advising and investment returns (Jackson et al., 2012), VCs have an incentive to reduce the cognitive load on their partners and make fewer investments. While such trade-off is well established, it is interesting that the number of investments per general partner varies highly among VCs. For instance, in the data set collected by Jackson and coauthors (2012), the number of ventures per partner varied from one to more than 21 with a median of close to seven. VCs that take on fewer investments per partner presumably are seeking to minimize the cognitive load on any given partner, allowing them to allocate more resources (financial and otherwise) to each venture (Bottazzi, Da Rin, & Hellmann, 2004). This concept links to research on a span of control (the number of direct reports), which often suggests that reducing the span of control can improve performance (e.g., Hansen & Wernerfelt, 1989). Our theoretical perspective is that a decrease in the load per partner will reduce resource constraints on the VC partners, which will provide an opportunity to deploy their existing expertise and offer advice in domains where they feel qualified in an effort to improve venture outcomes.
Hypothesis 5: A decrease in the VC PFCs per VC partner will be positively related to the likelihood of the VC to offer advice in familiar domains.
Method
Empirical Context and Data
Our research setting focuses on all 300 PFCs financed by 23 (of the population of 32) French VCs specializing in early-stage funding as of 2006. 2 Early-stage projects usually require greater attention and involvement than later-stage deals (Sapienza, 1992) and, as a result, represent a fruitful setting to study VCs’ advice giving. The context provides rich variance about problems (due to the nonlinear trajectory of most PFCs), proximity (since a portion of investments is done overseas), experience (based on the accumulation of ventures across VC portfolios), and constraints (due to the small size of most VCs). The data come from the following sources: questionnaires filled in by each general partner of each VC and follow-up interviews with VCs’ general partners and PFCs’ managers, VCs’ financial reports to their shareholders, and secondary sources, such as online platforms and databases (e.g., Societe.com, LinkedIn, Viadeo, VentureOne). We relied on mixed-methods research (Bryman, 2006; Plano Clark & Creswell, 2008) using exploratory sequential design (J. Greene, Caracelli, & Graham, 1989: 262) to build our questionnaire. This exploratory sequential design allowed us to validate that the value-adding activities carried out by VCs in the American context (e.g., MacMillan et al., 1989; Sapienza, 1992) were also carried out by VCs in France as well as augment it with other activities (e.g., internationalization of ventures, public grant applications) we identified via semistructured interviews. We also pretested our questionnaires with general partners to be certain that respondents understood terminology in exactly the same way. The questionnaires were completed during in-person visits with 22 of the 23 firms, which explains the high (72%) response rate. 3
By relying on different sources of data for our dependent and independent variables and cross-validating our measures from several different sources, we were able to maximize validity and avoid common-method biases (Podsakoff, MacKenzie, Lee, & Podsakoff, 2003; Podsakoff & Organ, 1986). For instance, one of the most frequent causes of common-method bias is when the same respondent provides both predictor and independent variables, where answers could be influenced by a consistency motif, implicit theories, social desirability, or leniency biases (Podsakoff et al., 2003). Information on VC advice giving (DV) came directly from VC partners via questionnaires. Partners provided information regarding the domains in which their VC had provided value-adding advice and helped establish networking to each portfolio company. 4 Since each survey was completed by the general partner who served on the PFC’s board, that is, the general partner most knowledgeable about the venture and the VC’s involvement (Sapienza, 1992), we were able to collect as complete information as possible about the scope of the VC’s involvement. To ensure that participants were interested and committed to providing accurate information, we guaranteed confidentiality and provided them a summary of the results. In addition to the information collected via questionnaires with regard to different domains, in our interviews with VCs, we also asked for specific examples and illustrations.
We collected information on all the independent variables (with the exception of information on ventures’ business model change) from archival financial reports of VCs to their limited partners. Indeed, we collected information on ventures’ business model change from various sources to minimize a single-source bias (questionnaire; archival records, such as financial reports or articles in the press; and—to address any discrepancies—interviews with the ventures’ CEO for clarification). Control variables were collected primarily from financial reports and in some cases were cross-validated with secondary sources (e.g., CEO replacement). All in all, by designing our study and collecting data on predictors and criteria from different sources, we were able to avoid common-method biases that are common when information on both dependent and independent variables come from the same survey respondent. Following the recommendation of Podsakoff and coauthors (2003: 897), “if the predictor and criterion variables can be measured from different sources, then we recommend that this be done. Additional statistical remedies could be used but in our view are probably unnecessary in these instances.”
Despite all the precaution taken in designing our study, we also investigated common-method variance by conducting a Harman one-factor test, which is the most widely used method in the academic literature for this purpose (Podsakoff & Organ, 1986). To do this, we entered all the variables into an exploratory factor analysis, principal component analysis with varimax rotation, and principal axis analysis with varimax rotation to determine the number of factors that are necessary to account for the variance in the variables. If a substantial amount of common-method variance is present, either a single factor would emerge or one general factor would account for the majority of the covariance among the variables. All these tests yielded six factors with an eigenvalue exceeding 1. These factors accounted for 65% of the variance. The factor with the greatest eigenvalue accounted for only 32% of the variance. Because no single factor emerged as a dominant factor accounting for most of the variance, the finding underlines that common-method variance is unlikely to be a problem in the data. We also took recommended steps to minimize potential social desirability and recall biases (Podsakoff et al., 2003), the details on which can be obtained from authors upon request.
We structured our data set at the PFC level—one observation per VC-backed PFC in the sample. Considering that one of the objectives of our study is to analyze the impact of experience of the VCs on the extent to which they offer advice on new activities within their investments, we restricted our sample to only those PFCs that were part of the VC portfolio at the time when we measured advice giving (148 current ventures). We use the other PFCs—all of which had exited by the time of our survey—to establish both VC experience and measures of the previous advice giving, as discussed below.
Measures
Dependent variables
Our dependent variables measure the number of new and familiar aspects of a given venture’s operations on which the VC gave advice or helped to establish network connections with the focal PFC. Our survey identified a universe of 18 business domains on which any of our VCs offered value added. These domains included advising on strategy, financing, marketing, product development, human resources, and internationalization, among others. To identify familiar domains for advice giving (domains where the VC typically gave advice to PFCs), we focused on the subset of investments from which the given VC had exited prior to making an investment in a focal PFC. Within the VC’s portfolio of exited ventures, we considered that a VC regularly advised on a specific aspect if the VC advised in this area to at least 33% of ventures (robust to other cutoff points discussed in the Robustness Checks section below). Thus, a VC with four previous exits who provided advice on financing to all four, on strategy to two, and on human resources to one would count finance and strategy as “familiar,” and anything else on which it provided advice in subsequent ventures—including human resources and the 15 domains where it had not previously provided any advice—as “novel.” The specific DVs used in the regression models are count measures (VC advice in familiar domains and VC advice in new domains).
Independent variables
Our first independent variable, PFC business model change, pertains to problem-driven search and is coded as 1 if a PFC substantially changed its initial business model and 0 otherwise. We defined a business model as “the resources, structures, and transactions with customers and suppliers that firm uses to create value by exploiting business opportunities.” We therefore defined a significant business model change as “substantial modifications in the way your firm is paid for the product or service it offers that could involve changes in products or services themselves, in organizational structure and in transactions with customers and suppliers.” Because business models and the related language are common among venture capitalists (George & Bock, 2011), we had little trouble ensuring that all participants understood this meaning, yet when pretesting our questionnaire, we still made sure that this understanding remained consistent. Our second demand-side variable, PFC location proximity change, is the difference between the proportion of the VC’s portfolio companies located in France (all of our VCs are in France) in the current portfolio versus the proportion in the firm’s exited venture portfolio. A positive difference implies a larger number of PFCs in the current portfolio being located within a geographic proximity to the focal VC (results for coding whether each specific venture is located in France or not are similar).
Our second set of independent variables relates to the supply side, such as the VCs’ experience and constraints. We coded VC exit experience as the count of previously invested ventures where the VC had already exited prior to its investment in the current venture. We coded VC % successful exits as the share of the VC’s previous investments that had resulted in a positive exit outcome, where positive outcomes included IPOs or trade sales with a positive valuation (above 0%). 5 Finally, VC load change is the difference between the number of PFCs per general partner in the previous (exited) portfolio and the number in the current portfolio.
Control variables
We sought to control for characteristics of the PFCs and VCs that may influence advice giving, based on prior research and our interviews with VC partners. At the PFC level, we controlled for PFC age at the time of VC investment (measured in years) and PFC sector (1 for information and communication technologies and 0 for biotechnology). 6 We also controlled for PFC valuation decline and coded it as 1 if the PFC’s valuation was positive (meaning the PFC created value compared to the initial investment), 2 if the valuation remained unchanged, and 3 if the PFC’s valuation declined from the initial investment. 7 Because the level of financial commitment of the VC influences the level of involvement with the focal PFCs (for instance, lead VCs are known to be more actively involved than followers), we controlled for the total amount of capital invested by a VC in a given portfolio company, VC investment. 8 Because VCs often syndicate with other VCs to share resources and effort, we control for the number of co-investors in each PFC, VC syndication. Because VCs vary in terms of advising activities and this may influence their cognitive availability for new activities, we controlled for the number of domains in which a VC intervened on a regular basis in its exited portfolio, VC # familiar domains. Because in our problem-driven set of hypotheses our goal is to understand the extent to which the “problemistic” nature (and not novelty, for instance) of business model change drives VCs’ advice giving on new and familiar domains, we needed to control for the extent to which a VC had dealt with these activities in the past. We, therefore, control for the proportion of exited ventures with business model change (% Exited PFCs w/business model change) to control for the extent to which VCs were engaged in this activity in the past. Because problems or anticipation of a subsequent change is often accompanied by a founder/CEO replacement in entrepreneurial ventures, we controlled for PFC CEO replacement (1 if replaced and 0 otherwise). In addition, time has been shown to be an impactful cognitive variable, so we controlled for the PFC time-to-exit forecast, which is the number of years that a VC predicts will be required until a potential exit. Finally, we also include year fixed effects to control for overall market trends.
Statistical Methods
Due to the fact that our data are ventures nested within VCs, we chose to use a multilevel model (xtmixed in Stata). This controls for the nested nature of the data and includes VC-level random effects to adjust for unmeasured factors. This is an appropriate model because, even though our dependent variables are count variables (counts of new and familiar domains), the variables are not overdispersed and the distributions are relatively normal. As robustness checks, we have tested our results with count models (both Poisson and negative binomial) and standard ordinary least squares regression, all providing consistent results. All results are clustered at the VC level.
Results
Descriptive statistics and correlations are presented in Table 1 and show a number of interesting descriptive findings. VCs offer more advice in familiar domains (7.6) than in new domains (0.8). While not shown, it is interesting to note that the total number of domains in which the VC gave advice in its previous versus current portfolios revealed that VCs typically gave more advice to ventures in their current portfolio than their exited portfolio. 9 This increase may come from advising on new topics or from consistently providing advice in each familiar area for every venture. Multicollinearity was not a problem, with most correlations below 0.3 and the variance inflation factors below 2.3, well below the maximum threshold of 10 suggested by Freund and Littell (1991).
Descriptive Statistics and Correlation
Note: N = 148; |correlation| >.14 significant at p < .05. VC = venture capital firm; PFC = portfolio company; ICT = information and communication technologies.
Centered variables used to reduce correlations, summary statistics reported uncentered.
Regression results for our hypotheses—problem-driven search (Hypothesis 1a, Hypothesis 1b), proximity (Hypothesis 2), experience volume (Hypothesis 3), experience valence (Hypothesis 4a, Hypothesis 4b), and constraint removal (Hypothesis 5)—are shown in Table 2. We first enter only control variables in Models 1 (for new domains) and 4 (for familiar domains). We add our independent variables corresponding to each hypothesis in Models 2 and 5, with nonlinear terms for experience added in Models 3 and 6. Based on the nonlinear results, we focus on Model 2 and Model 6 for our discussion. In general, we find support for problem-driven search. Specifically, we find that VC business model change is positively related to the VC’s engagement in new domains (Hypothesis 1a, Model 2, p < .01) and familiar ones (Hypothesis 1b, Model 6, p < .001). When a PFC experiences a major problem and undergoes a significant business model change, the VC will not only offer advice in familiar domains but also extend advice giving to new domains. Interestingly, the control variable for PFC CEO replacement—another potential signal of problems within a venture—shows a similar pattern in terms of advice giving in new and familiar domains, strengthening our suggestion of problem-driven search processes affecting both new and familiar domains. The coefficient for PFC location proximity change (Hypothesis 2) is also positive and significant (p < .05), supporting our suggestion that decreased geographic distance between VCs and PFCs leads to an increase in familiar, more routine advice-giving behavior (again, coding this variable as 1 if the PFC is in France and 0 otherwise produces qualitatively similar results).
Factors Affecting VC Advice Giving in New and Familiar Domains
Note: Standard errors are in parentheses. VC = venture capital firm; PFC = portfolio company; ICT = information and communication technologies.
p < .10.
p < .05.
p < .01.
p < .001.
In terms of experience, we find that both experience measures (volume and valence) have important relationships with advice giving. In Model 2, we see that a higher volume of experience is associated with less advice giving in new domains (p < .05), consistent with Hypothesis 3’s suggestion that experience would reinforce existing behaviors. The effect of experience on advice giving in familiar domains is somewhat mixed: negative in Model 5 and with a nonlinear, U-shaped effect in Model 6. To better understand the relationship, we plot the curves both for new and familiar domains in Figure 1, with the left-hand y-axis showing the number of domains in new areas and the right-hand y-axis the number of domains in familiar areas. The graph shows that firms tend to initially decrease their advice giving in familiar domains with higher levels of exit experience but later see the increases in advice giving that were predicted in Hypothesis 3. In general, these results are largely consistent with Hypothesis 3 that experience volume should be associated with decreased advice giving in new domains and increased advice giving in familiar domains, but the results add important nuance to our understanding of advice giving in familiar domains, as discussed in greater detail later.
Effect of Exit Experience on Advice Giving
In terms of the share of successful experience (valence), the results are somewhat clearer. Having a large share of successful exit experience leads to increases in advice giving in both new (Model 2, p < .001) and familiar (Model 6, p < .001) domains. In fact, in familiar domains, there is a positive main effect and squared term, though the shape of the curve is close to linear (see Figure 2). The effects in familiar domains are strongly consistent with Hypothesis 4a, which suggested that higher degrees of success would lead to increased emphasis on the VC’s existing array of familiar domains. The results are also consistent with Hypothesis 4b, which suggested that higher levels of successful experience would be associated with higher levels of advice giving in novel domains. In the theoretical discussion for Hypothesis 4b, we suggested that a positive relationship between successful experience and advising in novel domains could emerge through traditional slack search behavior but also through increased confidence in the VC’s abilities that may encourage it to expand the domains of its advising efforts.
Effect of Positive Exit Experience on Advice Giving
Finally, our results largely provide support for the theory that VCs extend advice to more familiar domains given lower time constraints. In line with our predictions, the coefficient for VC load change (Hypothesis 5) is positive and marginally significant (p < .10) in Model 6. As the load per VC partner goes down over time, the VC partners provide more consistent and higher levels of advice giving to their PFCs in the domains in which they have previous experience. Interestingly and consistent with our theory, we find that simply reducing constraints without either a sign of a problem or an increase in organizational slack has no effect on the cognitive exploration of new advice-giving domains.
The control variables generally perform as expected. VC time-to-exit forecast (reversed coding) is negative and significant (Model 2, p < .01) in explaining VC attention to new domains. The shorter the time is until the expected exit, the less likely the VC will engage in exploring new domains. Also, the larger the number of established advising domains (VC # familiar domains), the less likely the VC will engage in new domains (Model 2, p < .001) and the more likely the VC will devote attention to sharing knowledge in established domains (Model 6, p < .001). The former is consistent with the idea that VCs with multiple established domains for advice giving may have specialized knowledge and are less likely to stray beyond those domains. The VC is likely to reduce advising in some familiar domains for more mature ventures since we find that the coefficient for PFC age is negative and significant (Model 6, p < .05). 10
Robustness Checks
We also conducted additional robustness assessments. Our first concern was the relatively high correlations between variables. Our first approach to address this issue involved excluding individual variables with higher correlations (e.g., VC # familiar domains and % exited PFCs w/business model change), which produced similar results. The two experience variables of experience, VC exit experience and VC % successful exists, are also highly correlated (0.44). Excluding observations where the VC had no prior experience reduced the correlation to 0.26, suggesting that the correlation is largely driven by firms without experience. Second, when we tested the correlation between the volume and the valence of experience at the VC level (among all the 24 VCs), the correlation dropped to 0.37, and to 0.10 when we included only VCs with experience. Third, to be more certain about the implications of these VCs with no experience, we reproduced our full models using VCs with prior experience. The results are in columns 1 and 2 of Table 3 and show consistent patterns with our main results. Fourth, we normalized experience valence by volume and show these results in columns 3 and 4 of Table 3. The interaction terms are not significant, suggesting that normalizing may be less informative. Finally, we simply counted successes, failures, and their interaction and evaluated these results in columns 5 and 6 of Table 3. The results generally align with the main results reported in Table 2, emphasizing that the effects emerge more through negative versus positive experience. All these different analyses provide additional evidence that overall, our findings were robust to different specifications of these variables.
Robustness Tests
Note: VC = venture capital firm.
p < .10.
p < .05.
p < .01.
p < .001.
Next, we evaluated different dependent variables, including dichotomizing our dependent variable VC advice in new domains (equal to 1 if advising on any new business aspect, 0 otherwise), and creating a variable noting if the VC failed to give advice on a specific familiar domain to a given PFC. In addition, we have explored restricted samples, focusing only on VCs with prior exit experience or PFC-level observations where the VC had been on the board for at least 1, 2, or 3 years. We have also conducted robustness tests with lower cutoff points for what we considered as a familiar domain (e.g., 25% and 20%) as well as restricted our measure of novelty to those domains in which a VC advised only one new portfolio company and had not advised in this area in the past. The results of all these regressions (available from the authors) are generally consistent.
Finally, given that advice giving in both new and familiar domains is effectively a simultaneous decision for the VC partner, a simultaneous equation model would also be appropriate. This is less of a concern given that VCs generally increase the amount of advice giving that they give between their exited and current ventures, suggesting that they do not necessarily see their advice giving as bounded. There is also a positive correlation between advice giving in new and familiar domains in our data set, again reinforcing the lack of a trade-off between the two types of advice giving. Nevertheless, we wanted to consider a simultaneous model as an option, though this presented two challenges. First, a simultaneous model cannot be implemented with a nested, hierarchical model. Second, a simultaneous model requires a satisfactory instrument for each equation (W. Greene, 2003), which is a challenge in our models as most controls correlate with both dependent variables. While all instruments were weak, the best models used VC time-to-exit forecast as an exclusion restriction in the model VCF advice in familiar domains and used both PFC sector and PFC age as exclusions in the model VCF advice in new domains. Doing so produced only two changes from the models reported here: The negative and significant effect of VC exit experience on advice giving in new domains moved outside of significance (p = .17), and the positive and moderately significant effect of VC load change on advice giving in familiar domains becomes nonsignificant. Given the weak instruments, the change in model form, and the small sample size, these changes are not necessarily surprising.
Discussion
This study explored the determinants of managerial cognitive effort to new versus familiar domains in the context of the VCs’ advice giving to their portfolio companies. We built our theory around venture-related, demand-side factors—such as problem-driven (Greve, 2003) and proximity-driven advising—and VC-related, supply-side factors—such as organizational experience (Haunschild & Sullivan, 2002) and constraints (Penrose, 1959). We showed that each factor affected managerial cognitive effort but not necessarily in obvious ways. Specifically, we found that VCs increased advising their portfolio companies in new domains primarily when the venture displayed signs of problems or when the VC itself had slack resources accumulated from successful exits. Meanwhile, signs of problems within the venture also led to increased advising in familiar domains, suggesting that managers seek to solve problems both locally and with a more distant search. Yet, as VCs gained experience, they showed resistance to switching attention to new domains, even when having lower cognitive constraints (reducing constraints did not lead to more advice giving in new domains). Instead, having lower cognitive constraints and closer geographic proximity encouraged effort solely in familiar, well-understood domains. Meanwhile, negative prior experience reduced effort in both new and familiar domains, consistent with retreat from failure. Below, we explore how this study has important implications for the three streams of literature to which the study sought to contribute: managerial cognition, advising role of board members, and the VC–venture relationship. We also discuss the implications of contextual idiosyncrasies for the interpretation and extension of our findings to other settings and point out some limitations and opportunities for future research.
Cognitive Effort in Organizations
For the literature on managerial cognition and cognitive effort (Ocasio, 1997; Westphal, 1999), this study’s contribution is essentially twofold. It offers a novel theory and empirical evidence about the antecedents of managerial cognitive effort and identifies important differences between antecedents of cognitive effort directed toward familiar and novel domains. This is critical given that the ability to shift attention between familiar and new domains may be responsible for successful organizational adaptation and survival in turbulent environments. Our study uncovers specific sources of tension that organizations experience when allocating cognitive effort in different directions and therefore contributes to a deeper understanding of strategic organizational behavior (Eggers & Kaplan, 2009). Overall, while our research shows that organizations may switch attention to new domains when faced with challenges, the amount and nature of their prior experience may either support (when successful experience prevailed over negative experience) or restrain such effort (as a result of overall experience accumulation). Our findings also suggest that beliefs that cognitive and proximity-related factors restrain organizational adaptation efforts may be overstated in earlier research, since the decrease in both such constraints did not lead to explorative (but only to exploitative) efforts. We now discuss some of these findings and implications in more detail.
The most interesting and potentially challenging findings relate to problem-driven search and the role of experience on shifts in cognitive effort. First, our theory and results are consistent in showing that perceptions of problems within a venture lead to increased advice giving in both new and familiar domains. While this seems contradictory to prior research on problem-driven search (which typically suggests that problems lead to explorations in new domains, e.g., Greve, 2003), the original foundations of organizational cognition (e.g., Cyert & March, 1963) make it clear that organizations will respond to problems first by searching locally for a solution, turning to more distant potential options only when the local options are unsuccessful. As a result, problems should lead to an increase in the organizational effort to find solutions both locally and more distantly (potentially sequentially), which is highlighted by our findings. Much of the research in this tradition looked only at exploratory efforts while presuming that local search efforts must go down, which effectively treats exploration and exploitation as different ends of a continuum (see Lavie, Stettner, & Tushman, 2010). Instead, we suggest that future research looking at problem-driven search needs to consider both new and familiar efforts simultaneously, because the typical organizational response to problems may be to increase effort in both.
The second implication for research on organizational cognitive effort relates to the different findings for problems (e.g., business model change) and failures (e.g., unsuccessful exits). Our study shows that problems lead to problem-driven search as outlined above, while failures lead to decreased effort in new domains, consistent with a retreat from failure (Eggers, 2012). Thus, it appears that problems lead to an increase in search behavior, while failures lead to a decrease in search behavior. Why? We offer two potential explanations. One is simply that VCs expect failure and therefore do not substantially react to failures. A second is that while most behavioral research on problems and failures uses the words interchangeably (e.g., Greve, 2003), there are sometimes important differences between the two concepts in terms of the potential to repair the problem. For potentially repairable issues (problems), the results suggest that organizations will initiate search. For issues that are past the point of recovery (failures), our results suggest that retreat is more likely. Our findings thus underscore that it is important to consider how fixable a given problem may be in order to predict organizational response. Future research should consider and further examine this important distinction.
Third, our findings on the impact of the volume of experience turned out somewhat different from what we expected and deserve particular attention. Figure 1 showed that VCs tend to initially decrease their advice giving in familiar domains with higher levels of exit experience but later show the increases in advice giving predicted in our third hypothesis for firms with much more experience. This suggests that our hypothesized positive relationship might manifest in a sample of larger, more experienced VC firms (such as the U.S. context) but that very early on, experience tends to lead to retreat away from even familiar domains. This highlights that learning from experience is not an automatic linear process and that young organizations may face challenges with interpreting and effectively learning from their early experiences. Given that cognitive and behavioral research has greatly focused on larger established organizations, future research should dive deeper into understanding how organizations accumulate and learn from their experience at different stages of development.
Fourth, the interesting results in Figure 2 show that VCs with more successful experience increase their advice giving in both familiar and novel domains. While these results are consistent with reinforcement behavior (for familiar domains) and slack search (for novel domains), they are also consistent with patterns of superstitious learning based on increased confidence in the VC’s ability to affect venture outcomes. Superstitious learning (Levitt & March, 1988) occurs when confidence in one’s own competencies advances more rapidly than the rate of development of actual competence (Zollo, 2009: 897). Prior successes have been commonly seen as a source of overconfidence and self-attribution biases in managers’ perception, especially in settings where the high uncertainty of the outcome exists (Levitt & March, 1988). Given a high variety of factors contributing to the outcomes of VCs’ investments, VCs’ attribution of past successes to their own competencies and advising may undermine their learning of new knowledge. In other words, VCs that extend advice to new domains based on how successful they were in the past may not invest sufficient time and resources to acquire the level of competence high enough to add value to ventures in these domains. Future research—potentially in the setting of VC advice giving—could explore the extent to which the expansion of advice giving observed in our data is correlated (negatively) with venture outcomes to identify superstitious learning behaviors.
The Advising Role of the Board of Directors
The role of boards of directors in organizational survival and success has received considerable attention in the literature. The primary focus has been the monitoring function as a way to assure that managers perform their role as value maximizers for shareholders (e.g., Dalton, Hitt, Certo, & Dalton, 2007; Finkelstein, Hambrick, & Cannella, 2009; Tuggle, Sirmon, Reutzel, & Bierman, 2010) and the resource provision role (Pfeffer & Salancik, 1978) via interlocks that board members established with other organizations (Gulati & Westphal, 1999). While both monitoring and resource provision are essential, under increasingly unstable market conditions caused by innovation and increasing intensity of competition, the advising role of boards of directors has been recognized as of paramount importance to organizations navigating through untested waters (Carpenter & Westphal, 2001; Tuggle et al., 2010). Our study builds on this earlier research and reveals that both firm-centric and directors-centric factors influence not only the volume but also the direction of advice toward novel versus familiar domains and therefore should be considered concurrently. These findings lay fruitful background for future research examining performance consequences of boards’ advising.
Venture Capital
First, this study contributes to the literature on the benefits of VCs to new ventures (Hsu, 2006) and provides a novel explanation regarding mixed research findings in terms of the impact of such activities on performance. Our study is the first, to the best of our knowledge, to illustrate that VCs have knowledge domains in which they regularly provide advice and expand advice giving to new domains. This suggests that future research exploring the performance effect of VC advising should fully consider whether it builds on domains of strength for the VC or draws on less tested, newly acquired knowledge. Given inherent experiential learning differences between advising in novel and familiar domains together with the distinct motivational factors that we found, our research identified some potential sources of the mixed performance results of VC advising that future research should further explore.
One of the most intriguing contributions of our research is our finding of the impact of ventures’ problems on the direction of VC advising. The fact that we find that a venture’s problems increases the VC’s effort toward both new and familiar domains contrasts with a view of VCs as rational investors. Rational VCs should devote more time to their successful investments and withdraw from less successful ones in order to realize the returns necessary to make the fund a success (Gifford, 1997). When combined with our “non-finding” in terms of the impact of venture valuation decline (our control variable) on the VCs’ effort, we see several potential explanations of such VCs’ behavior. First, it is possible that VCs respond to problems by increasing effort only up to a certain point, without substantially compromising the amount of effort they could devote to well-performing ventures. Second, given how central a successful business model is for young ventures (Zott & Amit, 2007), VCs may perceive such effort as instrumental in bringing ventures’ performance back on track (Sahlman, 1990).
Limitations and Future Research Directions
This study also contains several limitations that point to opportunities for future research. First, it is possible that idiosyncrasies of our empirical setting might influence the study’s results. For instance, the VCs studied here are primarily smaller, younger, and less experienced than many of the prominent VCs in the United States. More experienced VCs may manage the cognitive load of partners, limiting the applicability of our theory around such constraints to all VC settings. Future work should explore the extent to which these findings do—and do not—translate to larger, more experienced VCs. Nevertheless, we think that our findings may resonate well with angel investors and angel funds as well as VCs in emerging markets. These investors contribute value-adding services in addition to providing financial capital to their portfolio investments. It is, therefore, possible that these investors will experience similar challenges (e.g., superstitious learning) with searching for knowledge and expertise that will be new to them.
Second, there are important caveats around generalizing our findings to more traditional organizational contexts. VCs have been described as more involved in advising than boards in many other contexts, making it a particularly appealing yet more idiosyncratic context in which to study board advising. In addition, when considering the role of problems and failures, it is important to recognize that VC partners expect more of their investments to result in failure than, for example, division managers in large existing firms. Thus, the dynamics of learning from failure and negative feedback may be different in other settings.
Third, while we built our theory on advice giving from the perspective of organizational cognitive effort, we recognize that we were able to measure only the output of potential effort but not the effort per se. This is an important limitation given that managers and organizations may vary in their absorptive capacity (Levinthal & March, 1993) to acquire and assimilate new knowledge (Ben-Oz & Greve, 2015). While it is reasonable to expect substantial effort behind value-adding advising, future research should try to measure more fully the process of advice giving and the extent to which its different aspects (e.g., knowledge accumulation, update, transfer, etc.) require cognitive effort from directors (e.g., VCs). In a similar fashion, even though our variable PFC business model change is highly relevant for our study, future research should incorporate different measures of problems and performance outcomes to gain a more holistic representation of organizational performance.
Fourth, our study is predicated on an ability to distinguish between novel and familiar domains, something that may not be feasible in all settings or intuitive to managers. While we agree that this point is an important boundary condition, we believe that the distinction between new and familiar areas of cognitive effort or knowledge is relatively intuitive and likely to be applicable in most, though not all, settings.
Finally, our main results are largely correlational. The accumulation of successful experience and the changing of a business model are not exogenously determined factors, and thus some caution must be used in interpreting our results. In addition, interpreting the effect of (for example) successful experience as a measure of revealed quality of a high-capability VC does not mitigate the fact that the VC itself needed to accumulate that successful experience before expanding the boundaries of its effort. The precaution that we took in our data sampling and analysis, combined with supplementary semistructured interviews with VC partners, gives confidence that this limitation should have no substantial impact on our findings. Future research on the antecedents of effort could focus on research designs that further limit the effect of endogeneity.
Practical Implications and Conclusions
Our study has practical implications for entrepreneurs raising funding from VCs. Establishing trust in a VC–venture relationship may be an important pathway to discussions (Sapienza & Korsgaard, 1996) and openness to soliciting a “third opinion” when a venture requires advice in domains that are beyond the focal VC’s expertise. For instance, we find that VCs advise both in new and familiar domains their PFCs that experience problems and undergo a substantial business model change. Because such business model changes are highly risky and may influence ventures’ survival and success, using different lenses when considering VC advice in new and familiar domains may offer valuable insight into the quality of the VC’s advising. Also, ventures that raise their funding from a foreign VC should be aware that the extent to which they could benefit from the VC’s advice is likely to be limited, since such VCs are likely to focus their effort on ventures located within a geographic proximity. Anecdotal evidence suggests that ventures should strive to raise funding from experienced VCs as a way to receive more specialized and valuable advice as they move forward. Our results point out that such biases as overconfidence and superstitious learning may interfere with the ability of experienced VCs to offer advice.
Overall, this study investigates the antecedents of managerial cognitive effort toward new and familiar domains by assessing the extent to which current problems, prior experience, and alleviation of time and resource constraints affect the propensity of VC managers to offer advice in the novel and familiar aspects to their portfolio companies. The theory and empirical findings contribute both to literature on the antecedents of managerial cognition and to literature on value added by VC firms. We also hope that our findings helped identify and direct scholars toward a number of potentially fruitful new avenues for future research.
Footnotes
Acknowledgements
This article was accepted under the editorship of Patrick M. Wright. Authors are listed in alphabetical order and contributed equally to this project. The authors wish to thank seminar participants at NYU Stern and Stanford University; conference attendees at the Academy of Management, the Babson Entrepreneurship Research Conference, and the Strategic Management Society; and Jonathan Arthurs, Rodolphe Durand, Nilanjana Dutt, Manuela Hoehn-Weiss, Sarah Kaplan, Sucheta Nadkarni, Ed Roberts, and Jung-hyun Suh for helpful comments on prior versions of this article. A condensed version of this article appears in the Academy of Management Best Paper Proceedings (2014) and the Frontiers of Entrepreneurship Research BCERC Proceedings (2015).