Observational Learning in Networks of Competition: How structures of attention among rivals can bring interpretive advantage

Abstract

Much of social network analysis has focused on learning in communication networks among collaborators in which actors can make direct inquiries to seek clarification about alters’ behavior or views. But such inquiries are typically not possible among rivals. Learning among rivals occurs primarily in observational networks in which actors must make inferences of the logics guiding their competitors’ behavior in markets. What promotes interpretive advantage in these networks of observation? We combine multimarket competition theory and structural hole theory to highlight the benefits of multiple exposure to disconnected competitors. In network-analytic terms we suggest that competitors’ interpretive advantage lies in non-redundant dyadic closure, especially when dealing with uncertain market niches. Dyadic closure, measuring ego’s exposure to her direct competitors in multiple markets, increases the ability to interpret competitors’ observed behavior. Redundancy, measuring the extent to which ego’s competitors are exposed to each other, reduces the diversity of views to which ego is exposed and hence the capacity to cope with uncertainty. We test our hypothesis by analyzing the network of competition created by securities analysts and the stocks they cover. We find that estimates issued by an analyst with multiple exposures to disconnected competitors are more accurate when confronted by more challenging, high risk, high reward, volatile stocks. Shifting the focus from direct social ties to the cognitive ties that link actors based on the objects, problems, or issues to which they pay attention, we develop a new approach to network analysis. Observation networks, we argue, operate neither as pipes nor as prisms but can be better conceived as scopes.

Keywords

attention networks closure market multimarket competition observational learning securities analysts social network analysis structural holes uncertainty

Departing from the atomistically calculating actor caricatured in the neoclassical economics framework, organization studies have exposed how individuals’ cognition is shaped by social structures (Kaplan, 2011; Porac, Thomas, & Baden-Fuller, 1989). Network scholars, in particular, have shown that informal collaboration and communication structures within organizations can facilitate learning and overcome individual bounded rationality. In communication networks among collaborators, researchers have emphasized the benefit of brokerage (or lack of triadic closure). Actors who interact with unconnected collaborators have an advantage as they can rely on diverse information sources and bring together different perspectives to tackle a focal problem (Burt, 2004).

What network scholars have not yet fully appreciated is that organizational actors operate in a broader competitive field with other organizations from which they can also learn (Huygens, Van Den Bosch, Volberda, & Baden-Fuller, 2001). Such learning from competitors differs from learning from collaborators in several aspects. First, instead of being driven by a common goal and mutual trust, learning from competitors is performed under the reciprocal wish to conceal information. Second, instead of being based on direct, face-to-face information exchange, learning from competitors rests on interpreting the observed behavior and inferring the strategy and logic that guide it (Greve, 2005). Because of the obstacles posed by such conditions, observational learning among competitors is often biased by misleading signals and subject to easy misinterpretation and can, therefore, ultimately translate into superstitious learning (Denrell, 2003; Levinthal & March, 1993).

The narrow bandwidth of observational learning suggests that the network structures that facilitate learning among competitors can differ from those that enable learning among collaborators. Whereas actors striving to achieve a common goal can leverage (relatively) strong ties with collaborators to gain access to diverse sources with relatively little effort, rivals have only limited access to their competitors’ knowledge. Multimarket-contact theory suggests that such access magnifies with the degree of competitive overlap in multiple markets (Jayachandran, Gimeno, & Varadarajan, 1999). As two rivals compete in a higher proportion of their respective market niches, they mutually recognize greater competitive interdependence (Gimeno & Woo, 1996). By heightening awareness and attentiveness between rivals, multimarket contacts create the condition for developing “competitive acumen”—the ability to understand a competitor’s perceptions and see things from a rival’s perspective (Tsai, Su, & Chen, 2011). A better understanding of competitors’ behavior can yield superior performance in markets because it makes it possible to more accurately estimate and preempt the retaliation cost of attacking them (Boeker, Goodstein, Stephan, & Murmann, 1997).

Our study provides a different insight into the learning benefits of multiple-market contact. It shifts the focus from learning about competitors’ likelihood to retaliate in a market to learning from competitors about that market. We focus, in particular, on the structure of observations among competitors that allows for superior interpretation of markets characterized by high uncertainty. It is under conditions of uncertainty, we argue, that attention to rivals is accentuated, and ego’s position in networks of observation becomes most salient.

To construct our theory, we leverage research exposing the countervailing effects of heedfulness to others’ interpretations in making sense of ambiguous, unknown, or novel situations (Gavetti & Warglien, 2015). On one hand, we reckon that multimarket exposure between competitors broadens the window to each other’s knowledge and allows for a superior understanding of each other’s interpretations. On the other hand, however, we believe that such exposure can reverberate interpretive noise and, ultimately, trigger a collective cognitive lock-in to biased interpretive models (Beunza & Stark, 2012; MacKenzie, 2011). The danger of cognitive interdependence prompted by such reflective modeling can be particularly calamitous when markets evolve rapidly and unexpectedly. In these uncertain scenarios, competition requires intelligence that breaks out of cognitive inertia and manages substantial ambiguity by creating “alternative representations,” “generative analogies,” and “appropriate recategorizations” (Tripsas & Gavetti, 2000; Gavetti, Levinthal, & Rivkin, 2005). As structural hole theory suggests (Burt, 1992, 2004), such conditions are met when actors enjoy exposure to non-redundant competitors.

The two processes—invoked by multimarket-competition theory and structural-hole theory, respectively—might, at first, appear to be antithetic. We show that they can be reconciled in specific network structures of competition characterized by non-redundant dyadic closure. Dyadic closure − capturing whether two actors can observe each other in multiple markets − allows for cycles of interpretation, enabling more in-depth vicarious learning. In a stylized example, we expect that a focal actor (ego) can draw better inferences about the logic that guides the interpretation of its competitor (j) on market 1 if she could observe the interpretation of j also on market 2. But we argue that the interpretive benefits of such dyadic closure are likely to bring a competitive advantage when the competitors that ego observes are not “redundant.” If ego can observe j and k in multiple but diverse markets, ego will benefit from the advantages of triadic openness. That is, when observing j and k, ego will be exposed to non-redundant, heterogeneous perspectives. It is the combination of such diversity of views (from open triads) and in-depth access to the views of one’s immediate rivals (through closed dyads) that provides the interpretive capacity that is necessary to cope with markets characterized by high uncertainty.

In the subsequent section, we outline the conceptual and methodological tools for moving from the analysis of social networks of interaction to the analysis of networks of observation. Here we define our terms, specify our research question, and formulate our hypothesis about how networks of competition shape the cognitive frames needed to interpret uncertain situations effectively. To test this hypothesis, we study the setting of securities analysts. Securities analysts compete to make timely judgments on situations that vary in uncertainty (i.e., they provide earnings estimates on firms with varying stock return volatility). They do so while embedded in networks of competition built around the stocks they cover. Our analysis of 430,384 earnings per share estimates issued by 11,259 analysts in the 1993–2007 period shows that the forecasts issued by analysts located in a network of competition characterized by closed dyads and open triads are more accurate when confronted by more challenging, high risk, high reward, volatile stocks.

We conclude by leveraging our results to offer a novel perspective on network research: networks can be seen not only as pipes disseminating information between collaborators or as prisms refracting identity to observers (Podolny, 2001). They can also be conceived as scopes sharpening or blurring vision among competitors observing each other.

Dyadic Closure: From learning about a competitor to learning from competitors about a market

Multimarket competition theory suggests that when firms “encounter each other in a considerable number of markets, [t]he multiplicity of their contact may blunt the edge of their competition” (Edwards, 1955; as quoted by Bernheim & Whinston, 1990, p. 1). Multimarket contact offers a firm attacked by a competitor in a market the opportunity to retaliate in the multiple markets in which the firms simultaneously compete. Anticipating the cost of a potentially widespread retaliation, firms that compete in several markets are more likely to engage in mutual forbearance (Baum & Korn, 1996, 1999; Bowers, Greve, Mitsuhashi, & Baum, 2014; Gimeno, 1999). More specifically, multimarket competition enables superior learning of the competitor’s motivation and competence to respond to an attack (Chen, Su, & Tsai, 2007). By improving the estimate of the likelihood of retaliation, multimarket contacts allow a focal firm the opportunity to calibrate the optimal intensity of attacking a rival in a given market without triggering a counterattack in other markets.

Although the literature has examined how multipoint competition opens a privileged window to the competitor strategy that is useful to understand the likelihood of reprisal to a competitive attack, it has not fully explored whether (and how) such a window can offer a better view of the particular market in which firms compete. To compete successfully in markets, firms must not only preempt offensive actions and reactions of their competitors, they must also achieve a deep understanding of the specific market in which their competition occurs. By observing its competitors, a firm might not only learn about them but also learn from them about the particular market.

The competitive setting we study − the setting of investment advice by securities analysts − allows us to explore this particular type of learning from a privileged standpoint. In many aspects, the competition of analysts parallels that of firms. Like international or diversified firms, analysts compete in multiple market niches (i.e., the stocks they cover). Like firms investing in various markets (e.g., entering multiple countries, launching diversified products, or targeting different segments), analysts must make accurate estimates about the prospects of each market. Finally, like firms, analysts can observe and interpret their competitors’ behavior and statements on those markets to refine their forecasts of each market’s profitability. The question we focus on is whether specific network structures of observation can bring an informational advantage to analysts by facilitating the learning from competitors that is needed to make sense of uncertain markets.

To answer this question, we should first ascertain the difference between observational learning and the explicit learning that can occur through direct formal instruction or one’s own direct experience (Bandura, 1977). A key difference is that, unlike interlocutors who can request elucidations to deepen understanding of their direct contacts’ knowledge, observers must rely on their own interpretation of the observed behavior. Learning by observing requires making inferences about the underlying logic that guides the observed behavior to generate structured, causal knowledge of the social world (Meltzoff, Waismeyer, & Gopnik, 2012). This inductive process of abstraction is facilitated by the analogical reasoning that unfolds by comparing observed behavior in multiple scenarios (Gentner, 2010; Goodman, Ullman, & Tenenbaum, 2011). It is through repeated observation of the performance of a comparable pattern/behavior across different situations that we learn the underlying logic that guides that action and thereby make inferences about the appropriate behavior to adopt also in cases we have not seen before (Walker & Gopnik, 2014).

The opportunity to make associations with behaviors observed in multiple contexts is critical among competing firms. Each firm has a strong incentive to prevent access to its knowledge since a sustainable competitive advantage lies in “causal ambiguity” and imitation barriers (Reed & DeFillippi, 1990). When trying to decode their competitors’ competencies by observing their behaviors, firms tend to construct an imperfect and incomplete account (Jayachandran et al., 1999). This nebulous understanding is most likely to occur when firms can only observe their competitors in one market niche only, especially when their attention is distracted by other market niches in which they compete with different actors. Observation in one market alone is exposure to behavior within a specific context. Contextually bound, this exposure offers a limited, blurred vision of the competitor’s logic.¹

Observation of a competitor in a greater number of market niches, by contrast, enlarges the window to the competitor’s behavior and provides opportunities to sharpen the understanding of the interpretive schema that underlies his strategy and guides his behavior (Tsai et al., 2011). Multimarket competition yields observational loops of recursive market interpretations. An actor (hereafter ego) who competes with another in two (or more) market niches can draw inferences from her competitor’s interpretations in one niche to make sense of the competitor’s interpretation of another. In turn, the competitor can draw inferences from the interpretation of ego in the latter market to make sense of ego’s interpretation in the first, and so on. We refer to these loops of recursive market interpretations as dyadic closure. Through the recursive loops of such closure, competitors can develop a superior interpretive capacity of each other. Each can vicariously learn what her competitor learned in the markets they cover.

The role of triadic closure and niche uncertainty

Even if dyadic closure facilitates vicarious learning from a competitor, it is unlikely by itself to give a competitive edge. The observational loops delineated by dyadic closure are likely to increase the ability to learn more quickly and deeply from a competitor. But, this learning will be mutual. The speed of mutual interpretive adjustment might not always be an asset (and, in some cases, it can be a liability) since it carries the risk, for both competitors forming the dyad, of becoming entrapped in a shared, univocal, and unchanging set of beliefs (Friedkin & Johnsen, 2011). To escape this spiral of beliefs convergence with a competitor, ego must be able to leverage exposure to another competitor with different, non-redundant beliefs. In networks, this exposure is granted by open triads.

The idea that open triads are needed to bring access to diverse views has been most fully elaborated by Ronald Burt (1992) in his concept of “structural holes” and an analytic method for identifying “constraint.” Drawing on the insight of Granovetter (1973) about “the strength of weak ties,” Burt noted that the limitations of closure create opportunities for actors who can fill the gaps (the structural holes) in network space. Actors who are embedded in a closed network lack access to interpretive diversity (Burt, 1992). Such access to diverse logic would be granted to actors who span across structural holes. By filling a network gap and occupying a structural hole position (i.e., by connecting to two actors who are not connected), ego can leverage a brokerage position and exposure to non-redundant alters to achieve a competitive advantage (Burt, 2004).

An actor can enjoy triadic openness in a network of competition when her competitors are not dyadically closed (i.e., when her competitors do not have multimarket contacts). From the perspective of ego, dyadic closure between her competitors is disadvantageous. When her competitors have multimarket contacts, they have greater opportunities to learn from each other (as well as from the same market niches they serve). For ego, such mutual and shared learning among alters would come at the price of limited diversity. In this scenario, what ego can learn from a competitor is likely redundant with what she can learn from the other. This redundancy will be proportionally reduced as ego’s competitors’ multimarket contacts diminish. Through their diverse experiences (formed in different, non-overlapping markets) and lack of mutual influence, ego’s competitors will be more likely to develop different cognitive frames.

However, just as we argued that ego’s dyadic closure with her competitors alone is not sufficient, so too, we argue that triadic openness itself is not sufficient. If ego lacks multimarket contacts with her competitors, she will be able to leverage only casual access to the views of the competitor, regardless of how diverse these are. Such casual access to (or lack of familiarity with) other competitors’ knowledge does not foster the development of the interpretive capacity needed to learn from them.

To be open to others’ views, we argue, actors must be closed in the network structure of their competitive “ties” (Dobusch, Dobusch, & Müller-Seitz, 2019). However, for such views to be functional for ego’s competitive advantage, they must be diverse. Thus, the question is how ego can benefit from diverse competitors’ views, and at the same time, be closed in the observational structures based on patterns of her competition with them.

In Figure 1, we depict for illustrative purposes three ideal-typical configurations that a focal actor can face while assessing a focal market niche. Top nodes in the figure (i.e., circles) represent competitors; bottom nodes (i.e., squares) are market niches. In all structures of competition, each actor focuses on three markets. What changes from the perspective of the focal actor is the opportunity to leverage loops of dyadic closure with each of their competitors and their redundancy.

Figures 1a and 1b illustrate the two ideal-typical configurations characterized by full openness and full closure, respectively. In the fully-open network structure (Figure 1a), ego encounters each competitor in the focal market only, and these competitors do not encounter each other in other markets. In this scenario of low dyadic and triadic closure, when assessing the focal market ego will benefit from exposure to diverse knowledge, knowledge that is developed by competitors who focus on distinct markets. Yet, lacking the opportunity to confront their interpretations in multiple markets, ego’s understanding of each competitor’s interpretation of the focal market will be limited. By contrast, in the fully-closed (both dyadic and triadic) structure (Figure 1b), ego is part of a cohesive attention cluster where all competitors focus on the same markets. In this scenario, the mutual understanding among ego and her competitors will be high, but the risk is that the echo of this mutual understanding will come at the cost of low knowledge diversity.

In between these two extreme ideal-typical cases, Figure 1c illustrates the position characterized by non-redundant dyadic closure (i.e., relatively high dyadic closure and low triadic closure). In this position, ego can see one competitor in one market and the other competitor in another market, while the two competitors are excluded from this opportunity. Because of the opportunity to observe them in other markets, ego will enjoy a privileged viewpoint when making sense of her competitors’ interpretation of the focal market. At the same time, because their attention is channeled towards different markets, ego’s competitors will approach the focal market with different perspectives, thus exposing ego to diverse views. Thus, we expect that by occupying a position rich in non-redundant dyadic closure, ego can benefit from an in-depth understanding of cognitively diverse competitors. When making her interpretations, she is best positioned to access diverse information, keep multiple scenarios in play, process complex information, and make timely judgments.

H1: Ego’s non-redundant dyadic closure is positively associated to interpretive accuracy.

Focusing on structural patterns of attention among competitors, we identified which position in a network of observation offers the opportunity to enhance mutual understanding and, at the same time, access interpretive diversity.

We now propose that to materialize the benefits provided by structure requires agency. To exploit their network, actors need to explore. They must deliberately activate their network ties to gather information. In a network of observation among competitors, this exploration implies paying careful attention to rivals and drawing inferences from their behaviors and interpretations.

Alertness and perusal of competitors’ views, however, constantly fluctuates. It increases or decreases as the problem that the actor faces needs closer or more distant search. This oscillation in competitor intelligence, we suggest, is punctuated by uncertainty in the focal market. Actors are subject to inertial behavior and engage in cognitive search only when the problem at hand requires exploring new solutions (Greve & Taylor, 2000). When the market follows expected trends (i.e., in situations of low market uncertainty), actors are likely to exploit their existing and consolidated knowledge in a routine fashion (Nelson & Winter, 1982). But this knowledge gets devalued when market uncertainty increases, and unforeseen market patterns demand scrutiny and novel sense-making (Sorenson, 2003). It is then that actors must initiate distant search processes of exploration that break out of their cognitive inertia. Research has shown that, in this search, actors open communication channels with their strong social ties in other organizations (Ozmel, Yavuz, Trombley, & Gulati, 2020) as well as weak ties with “competitors” (Botelho, 2018). We emphasize that they are also likely to increase their attention to their competitors’ views and interpretations.

Hence, we propose that the perspective offered by an actor’s position in the observation network will be more critical when market uncertainty is high. Moreover, we argue that it is the position of non-redundant dyadic closure that will prove beneficial in these uncertain contexts.

As Sorenson (2003) argues, market uncertainty not only devalues existing knowledge. It also introduces noise in the learning process. Most research assumes that interpreting this market noise needs leveraging diverse knowledge (Burt, 2004). By integrating diverse views, actors are more likely to make sense of situations that can change rapidly and follow unpredictable paths. However, exposure to diverse views in an uncertain scenario can also translate into a cacophony, as the abundance of dissonant messages amplifies the noise of market uncertainty.

For this reason, recent research in organizational studies has highlighted the importance of “buffers” when coping with market uncertainty. To be generative, dissonance needs orchestration. So far, the focus has been on organizational buffers that facilitate coordination among employees and effective integration of diverse knowledge. For example, Sorenson (2003) has highlighted the benefits of vertical integration as a buffer of external market uncertainty. More recently, He, von Krogh, and Sirén (2022) have pointed to the buffering of formal hierarchy and shown how it allows mitigating conflicts that can emerge when integrating diverse knowledge in the face of uncertainty. We echo this research but move the focus from internal organizational structures to external observational structures. We suggest that dyadic closure functions as a buffer of market uncertainty in that it reduces confusion in understanding competitors’ views. The ability to interpret these views quickly becomes key in conditions of uncertainty.

Nonetheless, dyadic closure alone, as we argued, will not be sufficient. Greater heedfulness to competitors allows for superior understanding, but it “can also generate conformity pressures that induce agents to give too much weight to others’ interpretations, even if erroneous, thereby potentially degrading interpretive performance” (Gavetti & Warglien, 2015, p. 1263). To break the vicious cycle of resonant messages circulating within the loops of dyadic closure, actors must also be exposed to diverse competitors’ views.

In summary, we posit that it is when coping with market niches with high uncertainty through redundant exposure (through high dyadic closure) to non-redundant actors (through low triadic closure) that actors become open to embracing alternative logics that free them from the lock-in of their consolidated models, break out of cognitive inertia, and manage substantial ambiguity. By activating deep access to diverse competitors’ views, actors find the proper context to interpret uncertain situations accurately.

H2: Market niche’s uncertainty positively moderates the effect of ego’s non-redundant dyadic closure on interpretive accuracy.

Data and Methods

Data and setting

We study how an actor’s location in a network of competition affects her ability to interpret situations of high uncertainty in the context of securities analysts. This setting is advantageous for several reasons. First, among securities analysts, we can construct in a timely and objective fashion the (analyst-stock) network of competition (Zuckerman, 2004). Analysts are employed by brokerage houses to provide investment advice to the brokerage house’s clients (the investors) about securities that are listed on the stock exchange. Each of these securities represents a market niche where the analyst competes with other analysts (Bowers et al., 2014; Hong & Kacperczyk, 2010). Second, the securities analyst setting is suitable for studying how competition intensity opens vicarious learning opportunities. For each firm they cover, analysts write detailed reports containing their earnings estimates and a sophisticated analysis supporting them. Just as companies reverse-engineer their competitors’ products to understand how they can improve their own, analysts indicate that they read their competitors’ reports to refine their models. Finally, the stocks that analysts cover differ in their market price volatility, thus providing a setting where we can assess interpretive accuracy under conditions of different levels of uncertainty. The future price of stocks exhibiting high price volatility is notoriously challenging to predict. But it is precisely in such high risk conditions that high returns can be gained.

We collect analysts’ coverage and earnings estimates data through the Institutional Brokers’ Estimate System (IBES).² To avoid the rounding problem identified by Payne and Thomas (2003) with the adjusted file in IBES, we use the unadjusted file and bring all estimates on the same split-adjusted basis without rounding. Our final sample encompasses 430,384 analysts’ earnings estimates from 1993 to 2007. We complement IBES data with other information. Specifically, from the Center for Research in Security Prices (CRSP), we obtained information on stocks’ splits and the stock’s daily market returns, which are needed to compute the firm’s stock returns volatility. In addition, to gauge differences in status among securities analysts, we collected data on the Institutional Investor All-Star ranking, issued every year in October to designate the best analysts in each industry.

Dependent variable and models

Interpretive accuracy

We measure analysts’ interpretive accuracy with the error they make in predicting the end-of-year earnings of the firm they cover. More specifically, we measure the absolute forecast error as the absolute value of the difference between the expected earnings-per-share of a focal firm and the actual earnings announced by that firm at the end of the fiscal year. More formally:

A b s o l u t e F o r e c a s t E r r o r_{i, s} = | E P S_{i, s} - E P S_{s} |

where EPS_i,s is the last estimate issued by analyst i predicting the year earning of firm s before the earnings announcement, and EPS_s is actual earning reported by the firm. As a common practice in the finance literature, we exclude stale forecasts issued 365 days before the earnings announcement.³ The resulting variable is non-negative positively skewed. We follow recent literature and test our hypotheses with Poisson estimations (Silva & Tenreyro 2006). Specifically, we use the PPMLHDFE Stata module for Poisson pseudo-likelihood regression with multiple levels of fixed effects (Correia, Guimarães, & Zylkin, 2020).⁴ We estimate models with analyst and firm-year fixed effects. Analyst fixed effects account for time-invariant characteristics of the analyst we cannot control for (such as innate abilities, IQ, education before joining the analyst’s profession, gender, etc.). Firm-Year fixed effects allow us to control for underlying attributes of the covered firm such as market size, past performance, diversification level, analysts’ competition around the stock (i.e., number of analysts covering it), as well as key attributes of the industry in which the firm competes (e.g., growth, concentration, etc.). Moreover, firm-year fixed effects allow us to estimate the effect of non-redundant dyadic closure within the same predictive task.

Independent variables

Analysts’ accuracy depends on their location in the network of competition. We construct the network of competition each calendar year based on the stock coverage of securities analysts. We thus consider that a tie between an analyst and a firm exists if the analyst issued an earnings estimate about that firm that year.

The network thus constructed, however, might be the result of analysts’ selection biases. These biases can raise endogeneity concerns if some unobserved analyst’s time-varying attribute we cannot control correlates with the hypothesized advantageous network position. We mitigate these concerns by implementing an instrumented network approach (Boehmke, Chyzh, & Thies, 2016).

In the first stage of this approach, we predict the likelihood of each observed network tie formation as a function of an instrumental variable. Precisely, we predict the probability that an analyst covers a firm (the formation of a tie in our network) because of a firm-specific characteristic that defines securities that analysts must cover: the firm’s inclusion in the Standard & Poor’s 500 index (S&P500) (Yu, 2008).

To estimate this probability, we need first to identify the ties that were not formed but were at risk of being created. In other words, for each stock covered by the focal analyst, we need to identify a matched stock the analyst did not cover but could have covered. We identify these peer stocks based on Hoberg and Phillips’ (2010) similarity scores⁵ and match each covered stock by an analyst with the closest non-covered stock in the business description space. Two firms with high similarity scores operate in the same business and are at risk of being covered by an analyst specializing in that industry. We use this matched sample to estimate the probability of covering any given stock as a function of the S&P500 listing and use the predicted probabilities from this first-stage model to weight each tie that composes the stock coverage network in the given year t.

The resulting network is a weighted network where each tie is weighted by the probability of being formed for reasons independent of the analyst’s endogenous decision. Stocks covered because of the inclusion in the S&P500 get a greater weight than stocks covered for the analyst’s likely endogenous decision, such as the decision to occupy a specific network position or follow certain competitors. As a robustness check, we compute non-redundant dyadic closure (described below) for each analyst in this instrumented weighted network.

Non-redundant dyadic closure

Our primary independent variable captures the extent to which the loops of dyadic closure are not redundant from the focal analyst’s viewpoint. To measure that variable, we need first to compute the cycles of dyadic closure between the focal actor and her competitors. We compute this measure in the observed-unweighted and (as a robustness check) in the instrumented-weighted network. In the observed-unweighted network, we count these loops based on the number of stocks that the focal analyst i covers with the other analysts covering the focal stock s. Specifically, the dyadic closure of analyst i while estimating earning of stock s corresponds to:

D y a d i c C l o s u r e_{i, s} = \sum_{j \in N (s)} \frac{(I_{i j}) * (I_{i j} - 1)}{2} i \neq j

where N(s) are the first-order neighbors of stock s (i.e., the other analysts who cover the focal stock), and I_ijis the number of stocks that i and j cover together. The number of four loops is a combinatorial function of the dyad’s stock coverage overlap.⁶ We normalize this number by the maximum number of loops in the coverage network that year so that it ranges between zero and one. In the instrumented-weighted network, we sum the weights of the ties that form the four-step closed loops, and as before, normalized by the maximum value in the network that year.

We then calculate the loops of dyadic closure that are not redundant from the focal analyst’s viewpoint. Loop redundancy occurs in the case of triadic closure, which is when the interpretive loop that i can leverage to refine her understanding of j overlaps with k. Our measure of non-redundant dyadic closure for analyst i and stock s is thus given by:

N o n - R e d u n d a n t D y a d i c C l o s u r e_{i, s} = \sum_{j \in N (s)} \sum_{k \in N (s)} {\frac{I_{i j, k} * (I_{i j, k} - 1)}{2}}_{i \neq j \neq k}

where I_ij,kis the number of stocks that i have in common with j but not with k . Thus, this measure counts the number of four-step loops between i and j that are not redundant with k.⁷ This measure requires at least three analysts covering the focal stock. For example, when applied to the network drawn in Figure 1b above, this measure would count zero non-redundant loops. Even if i can benefit from many loops of dyadic closure with all competitors she confronts on stock s, none of them is exclusive. Even if i is likely to understand each of his competitors’ interpretations of s deeply, she will not have any competitive advantage. By contrast, in the network depicted in Figure 1c, the focal analyst has two non-redundant loops of dyadic closure. Therefore, according to our measure, the configuration drawn in Figure 1c has the highest non-redundant closure. In the instrumented-weighted network, we sum the weights of the ties that form the non-redundant closed loops. As above, we normalize non-redundant dyadic closure with the maximum value of that year.

Niche uncertainty and stock price volatility

To capture the uncertainty of the market niche that the analyst is assessing, we use the firm’s stock-return volatility 30 days before the earnings announcement date. The greater the standard deviation of the stock return, the greater the uncertainty around the future earnings of the firm in the market, making the task of predicting firms’ earning more challenging.

Control variables

We control for several variables that might affect the ability of the analyst to issue accurate forecasts. Following the finance literature, we account for the experience of the analyst. We control for both the analyst’s professional experience and the specific experience acquired with the particular stock (Mikhail, Walther, & Willis, 1997). We measure analyst professional experience as the number of years that elapsed since the first estimate issued by the analysts in IBES and the beginning of the focal year in which we measure the analyst earning accuracy. Analyst stock-specific experience is the number of years from the first estimates issued in IBES by the analyst on the specific stock. We also control for the analyst status, as measured by whether she is listed as an All-Star analyst in the Institutional Investor magazine ranking (Phillips & Zuckerman, 2001) and the status of the brokerage house for which the analyst works, as measured by the number of All-Star analysts employed by the broker.

To control for herding behavior by the analyst, we include a forecast boldness variable, capturing the extent to which the estimate deviates from the standing consensus estimate (i.e., the average among the forecasts issued by other analysts before the focal analyst’s forecast).⁸ To account for the different timing of forecasts and for the greater challenge to make predictions that are more distant in time, we control for forecast horizon, measuring the number of days that elapse between the forecast date and the date of earning announcement by the firm.

To ensure that our results are not biased by other features of the analyst’s ego-network such as the number of direct and indirect ties (Ahuja, 2000), we control for the number of stocks covered by the analyst (i.e., her degree centrality or the number of direct ties) and the number of competitors she faces on these stocks (i.e., the number of second-order neighbors or number of indirect ties). To exclude the possibility that the effect of the network position could be driven by ‘connections’ to analysts who cover many stocks, we also control for the average number of stocks covered by the analyst’s competitors. To be more confident that the network position occupied by the analyst is not simply the result of covering different industries, we control for the number of industries covered by the focal analysts in the focal year.⁹

Findings

In Table 1, we report the descriptive statistics of our variables, and in Table 2, the correlation matrix.

Table 1.

Descriptive Statistics.

	mean	sd	min	max
Absolute forecast error	0.11	0.18	0.00	1.59
Dyadic closure	0.06	0.10	0.00	1.00
Non-redundant dyadic closure	0.03	0.06	0.00	1.00
Stock’s return volatility	0.03	0.02	0.00	0.67
Analyst professional experience	6.82	5.22	0.00	27.00
Analyst stock-specific experience	3.18	3.44	0.00	26.00
All-star analyst	0.12	0.32	0.00	1.00
Forecast horizon	116.91	87.57	1.00	365.00
Forecast boldness	0.07	0.11	0.00	0.79
No. of stocks covered	15.27	7.32	1.00	37.00
No. of competitors	77.66	39.49	1.00	395.00
Average No. of stocks covered by the analyst’s competitors	13.62	2.50	1.00	32.00
No. of industries	2.00	1.35	0.00	7.00
No. of all-star analysts	10.91	15.31	0.00	59.00
N	430,384

Table 2.

Correlation Matrix.

	1	2	3	4	5	6	7	8	9	10	11	12	13
Absolute forecast error	1.00
Dyadic closure	−.01
Non-redundant dyadic closure	−.03	.81
Stock’s return volatility	.08	−.16	−.12
Analyst professional experience	−.01	.14	.11	−.08
Analyst stock-specific experience	−.00	.28	.21	−.16	.56
All-star analyst	−.01	.22	.15	−.07	.20	.20
Forecast horizon	.20	−.10	−.09	.03	−.01	−.03	−.13
Forecast boldness	.58	.03	.01	.07	−.00	.02	.01	.16
No. of stocks covered	−.02	.37	.29	−.06	.27	.18	.16	−.04	−.03
No. of competitors	−.04	.10	.21	.10	.17	.07	.02	.00	−.04	.55
Average no. of stocks covered by the analyst’s competitors	−.02	.07	.00	−.11	.00	.02	−.01	.01	−.05	.24	.03
No. of industries	−.04	−.04	−.05	−.08	.20	.14	.03	.01	−.05	.33	.41	.12
No. of all-star analysts	.01	.19	.12	−.04	.11	.12	.44	−.04	.02	.02	−.09	−.11	−.06

In Table 3, we report the results of our statistical analysis estimating the effect of the analyst’s network position on the analyst’s accuracy. Model 1 reports our baseline results with control variables and the measures of dyadic closure only. In model 1a, we use the measure of dyadic closure calculated in the observed-unweighted network; in model 1b, the one computed in the instrumented-weighted network. We start by noting that the relation between volatility and the analyst’s forecast error cannot be estimated because of the firm-year fixed effects. In models without such fixed effects, the coefficient is positive and significant, thus suggesting that the volatile stocks are more challenging to predict: the higher the stock’s volatility, the higher the analyst’s error in predicting the firm’s earnings. However, as previously mentioned, it is precisely in these challenging situations where opportunities lie. In financial markets, stock return volatility is a proxy for risk (Goyal & Santa-Clara, 2003), and the higher the risk, the higher the potential return.

Table 3.

Poisson Pseudo-Likelihood Regression with Analyst and Firm-Year Fixed Effect Estimating the Effect of Non-Redundant Dyadic Closure on Analyst EPS Forecast Error (1993–2007).

	Model 1a	Model 1b	Model 2a	Model 2b	Model 3a	Model 3b
Constant	−2.397***	−2.373***	−2.398***	−2.372***	−2.398***	−2.373***
	(0.045)	(0.051)	(0.045)	(0.051)	(0.045)	(0.051)
Analyst professional experience	−0.006	−0.019**	−0.006	−0.019**	−0.006	−0.019**
	(0.005)	(0.006)	(0.005)	(0.006)	(0.005)	(0.006)
Analyst stock-specific experience	−0.001	−0.001	−0.001	−0.001	−0.001	−0.001
	(0.001)	(0.001)	(0.001)	(0.001)	(0.001)	(0.001)
All-star analyst	0.005	0.008	0.005	0.008	0.005	0.007
	(0.006)	(0.007)	(0.006)	(0.007)	(0.006)	(0.007)
Forecast horizon	0.003***	0.003***	0.003***	0.003***	0.003***	0.003***
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
No. of stocks covered	−0.001	−0.000	−0.001	−0.000	−0.001	−0.000
	(0.001)	(0.001)	(0.001)	(0.001)	(0.001)	(0.001)
No. of competitors	0.000**	0.000	0.000**	0.000	0.000**	0.000*
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
Average no. of stocks covered by the analyst’s competitors	0.002(0.002)	0.003(0.002)	0.002(0.002)	0.003(0.002)	0.002(0.002)	0.003(0.002)
No. of industries	0.001	0.001	0.001	0.001	0.001	0.001
	(0.002)	(0.002)	(0.002)	(0.002)	(0.002)	(0.002)
No. of all-star analysts	−0.001**	−0.001*	−0.001**	−0.000*	−0.001*	−0.000*
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
Forecast boldness	2.175***	2.196***	2.175***	2.196***	2.175***	2.196***
	(0.019)	(0.020)	(0.019)	(0.020)	(0.019)	(0.020)
Dyadic closure	−0.063*(0.032)	−0.031(0.033)
Non-redundant dyadic closure			−0.116*	−0.109*	0.124	0.132
			(0.049)	(0.050)	(0.084)	(0.085)
Stock’s return volatility × Non–redundant dyadic closure					−10.951***(3.094)	−10.699***(3.028)
Analyst fixed-effects	yes	yes	yes	yes	yes	yes
Firm-year fixed-effects	yes	yes	yes	yes	yes	yes
Log–likelihood	−104191.6173	−91660.6281	−104191.5528	−91660.4796	−104191.0542	−91659.9900
N	427,851	384,198	427,851	384,198	427,851	384,198

Standard errors in parentheses.

*p < 0.05, **p < 0.01, ***p < 0.001.

Turning to the study of network effects on analysts’ accuracy, in model 1a, we observe a negative coefficient between dyadic closure and our variable of interest. Because we estimate analysts’ forecast error, the negative coefficient indicates that analysts whose network of observation is rich in dyadic closure are more accurate: as dyadic closure increases, the analyst’s forecast error decreases. In model 1b, this effect is not significant. Models 2a and 2b introduce our measure of non-redundant dyadic closure while excluding dyadic closure. As before, this variable is negatively associated with the forecasting error, and significant when measured both in the observed and instrumented network. These findings support our first hypothesis suggesting that the advantage of dyadic closure arises when it offers exposure to non-redundant competitors. Models 2a and 2b, however, estimate the effect for an average value of stock volatility. We expect that the benefit of occupying a position rich in non-redundant dyadic closure will vary with volatility. Specifically, in H2, we hypothesized that it would increase in conditions of high uncertainty. We tested this hypothesis in models 3a and 3b, in which we introduced the interaction between our network measure and the stock’s market return volatility. We obtained a negative and significant coefficient for both models’ interaction effect between non-redundant closure and volatility. This result points to a more beneficial effect of the network configuration of high dyadic closure and low triadic closure when assessing volatile stocks. We depicted the marginal effect (at the mean value of other variables) estimated in model 3b in Figure 2. The figure shows the impact of non-redundant dyadic closure on analyst forecast error for relatively low (one standard deviation below the mean) and high (one standard deviation above the mean) uncertainty (as measured by past stock return volatility).

As Figure 2 shows, when there is high uncertainty around the stock’s future earnings, analysts can gain a significant vision advantage through non-redundant closure. In contrast, when the market is stable and predicting firms’ future returns is less challenging, non-redundant closure seems not to make a considerable difference.

Discussion

Our paper contributes to several streams of literature. First and foremost, our work contributes to the literature investigating the effect of multimarket competition on actors’ performance. When two competitors simultaneously compete in several markets, this literature acknowledges that the outcome of their choices mutually depends on each other (White, 1981). As a result of the increased competitive interdependence, actors become more prone to mutually forebear one another (Baum & Korn, 1996, 1999; Bowers et al., 2014; Gimeno, 1999) and more aware of their motivations and competencies to retaliate (Chen et al., 2007). Thus, multimarket competition can enhance an actor’s market performance by deterring a competitor from attacking and learning his disposition to respond to an attack.

Our work emphasizes that, through multimarket contacts, competitors also mutually expose each other to their views and ways of doing things on the markets they cover. By shifting the focus from learning about each other to learning from each other about those markets, we argue that simultaneous presence in multiple markets alone, albeit needed, is insufficient to bring superior performance. Multimarket contacts allow for loops of recursive interpretation that improve vicarious learning. But the cognitive interdependence of such reflective interpretations can sometimes be calamitous: when the competitors’ observations lack interpretive diversity, the resulting resonance (a collective cognitive lock-in) can lead to an “interpretative bubble” (Beunza & Stark, 2012; MacKenzie, 2011). To escape such a bubble, we suggest that actors must also be exposed to diverse interpretations from non-redundant competitors (i.e., competitors with limited multimarket contacts between themselves). By juxtaposing their contrasting views, ego can benefit from the deep understanding of her competitors without being entrapped in a spiral of convergent, unquestioned beliefs (Gavetti & Warglien, 2015). In networks of competition, we show, these benefits are achieved through relatively high dyadic closure and low triadic closure.

Studying such networks involved a shift from direct social ties to the cognitive ties that link actors based on the objects, problems, or issues to which they pay attention. Our paper developed this insight, which we formulate here. Networks of cognition among competitors are observation networks. Because these networks are formed by competitors linked by ties of indirect observation (rather than by collaborators linked by ties of direct communication), they cannot be conceived either as pipes or as prisms.

Both the pipes and prisms perspectives focus on direct ties of collaboration between producers. The former suggests that when producers build a collaboration tie, they create a “pipe” through which knowledge and information flow (Granovetter, 1973; Podolny, 2001). These pipes serve producers to cope with their own “ego-centric” uncertainty − the endogenous uncertainty related to their innovative and creative process (Reagans & Zuckerman, 2001; De Vaan, Stark, & Vedres, 2015). In the network as pipes perspective, “ego-centric” uncertainty is lessened by gaining access to diverse views. Typically, this condition is met when one’s direct contacts are not redundant, in other words, they do not interact with each other (Podolny, 2001, 2008).

When conceived as “prisms,” the social relationship between collaborators is interpreted through the eyes of third-party observers, such as consumers or investors (Podolny 1993, 2001; Stuart, Hoang, & Hybels, 1999). Notable in this perspective is that observers’ attention to the patterns of edges (the ties among the collaborators) allows them to reduce “alter-centric” uncertainty—the observer’s uncertainty about the collaborators’ quality. In the network-as-prisms view, such uncertainty is reduced by exposure to redundant messages that signal producers’ clear, robust identity that audiences evaluate¹⁰ (Zuckerman, 1999).

The network we studied is comprised of producers observing other producers (White, 1981). The uncertainty they cope with is the “market-centric” uncertainty of the niches they attend. Our proposition is that actors can better deal with that uncertainty through redundant exposure to non-redundant (diverse) views. Our argument builds on the fundamental differences that characterize learning through ties of communication among collaborators vs. ties of observations among competitors.

Building on the first difference in learning bandwidth between communication ties vs. observation ties, we illustrate a different process through which actors can grasp each other’s non-codified, tacit knowledge. As research in networks of collaboration has extensively illustrated (Hansen, 1999; Uzzi, 1996), codified knowledge among collaborators can be easily transferred via weak ties, whereas non-codified, tacit knowledge requires investments in strong ties: “The two-way interaction afforded by a strong tie is important for assimilating the non-codified knowledge because the recipient most likely does not acquire the knowledge completely during the first interaction with the recipient but needs multiple opportunities to assimilate it”(Hansen, 1999, p. 88; Polanyi, 1966). Competitors (such as, in our case, securities analysts) all have access to the codified knowledge embodied in each other’s products (their reports). To gain access to the tacit knowledge behind their artifacts, competitors, in the same way as collaborators, need multiple opportunities to assimilate it. Such opportunities among competitors are not realized by strengthening a direct personal tie but rather by enlarging the observational window to the competitor’s behavior by observing the application of the competitor’s tacit knowledge across different situations. In network-analytic terms, tacit knowledge among competitors is accessed not through strong ties but through dyadic closure.

As we described, ties among collaborators and those among competitors have a second crucial difference: they are not only of different types (attentional vs. social) and bandwidth (low vs. high), but they also serve different goals. Contrary to collaborators whose goal is to coordinate efforts for a common objective, competitors strive to gain access to the distinctive knowledge that would bring a competitive edge. Accordingly, we also argued that, whereas in networks of collaboration, the strong ties of trust and personal familiarity that are forged by repeated interaction might prove sufficient to enable the knowledge sharing that is needed for coordination, in networks of competition, dyadic closure alone will hardly give a competitive advantage. Dyadic closure will grant superior access to the competitor’s tacit knowledge but, in itself, will not provide performance superior to the competitor (more likely, it will foster convergence between the tacit knowledge of ego and her competitor). At the extreme, it can foster the formation of an investment community converging toward the same model (MacKenzie, 2011). For example, when a set of investors form cohesive ties of observations around the same set of stocks, they “should have relatively little trouble arriving at a common approach for deciding whether a piece of information is relevant” (Zuckerman, 2004, p. 413).

Much of the network analytic response in the literature to the question of how ego can leverage her network position to gain an advantage vis-a-vis others has pointed to brokerage, especially when it spans otherwise disconnected parts (or “structural holes”) of the network (Burt, 2004; Podolny, 2008). The argument for the benefit of brokerage often invoked in social network analysis is that the broker has a superior vision because, unconstrained, she can flexibly shift viewpoints (Burt, 2010). In networks of collaboration, a producer positioned in a structural hole position enjoys an advantageous perspective because she can leverage an unconstrained ability to shift her glance quickly from one direct contact’s viewpoint to another non-redundant one, gaining new information with each shift. In networks of competition, however, for the reasons described above, the bandwidth through which information and insights can flow will be considerably limited when compared to the bandwidth characterizing networks of collaboration.

Hence, as dyadic closure alone would not be sufficient to gain an advantageous position in networks of competition, so too brokerage alone will not be enough. In networks built around ties of observations among competitors, it is not sufficient to have sporadic access to alters’ views, regardless of how diverse. Weak, non-redundant observations of competitors might have little influence on ego and thereby leave her unconstrained. But, as sporadic, they are also shallow: they offer only a blurred vision of competitors’ knowledge. To leverage the diverse views of one’s competitors, to see through their eyes, we showed that one needs not only the lack of triadic closure but also the presence of dyadic closure.

Thus, our answer to the core question of which position gives a “vision advantage” to deal with market uncertainty builds on the third difference between networks of competition vs. networks of collaboration. Whereas networks of collaboration are formed around direct ties (referred to by network analysts as “one-mode” networks because they are comprised of one set of nodes − actor to actor), networks of observation are built on indirect ties (i.e., two-mode networks linking two set of nodes − actor to objects). Our conceptualization of networks of observation as two-mode allows us to engage with a key trade-off that has characterized network research from its inception, particularly in the pipes perspective characterizing one-mode networks of collaborations. As it is commonly invoked in that literature, the benefits offered by dyadic patterns typically contrast with those stemming from triadic structures. Whereas strong dyadic ties of personal familiarity offer a higher volume of information flow (Hansen, 1999; Uzzi, 1996), open triads provide higher diversity of information flow (Burt, 2004). Nevertheless, the possibilities of obtaining both benefits simultaneously are unlikely (hence, a trade-off) because the processes that promote strong dyadic ties are likely to produce triadic closure (Granovetter, 1973).

As Burt (1992) argued, however, the strong dyad vs. open triad trade-off is not absolute: tie weakness is a correlate, not a cause, of non-redundancy even if less frequent strong ties can also be formed with non-redundant contacts (Vedres, 2017). As Podolny (2001) observes, “In fact, controlling for the extent to which a tie serves as a bridge to distinctive sources of information, stronger ties are actually more beneficial than weak ties since they allow a greater volume of resources to move between actors” (p. 34). Most organizational studies have, nonetheless, discounted this possibility. This is not surprising because the main focus of organizational research has been in the intra-organizational contexts of co-workers sharing physical space or in the inter-organizational contexts of directors meeting in boardrooms face-to-face. In these contexts, dyads who build an emotional bond and a relationship of mutual trust (i.e., a strong tie) are unlikely to be dis-embedded from closed triads. Thus, to date, research has not systematically worked out the combinations of strong ties and triadic closure in formally structural terms. When investigating the combined benefits of strong ties and open networks, existing research has not examined triadic closure explicitly. Instead, it has looked at how strong ties can bring enhanced benefits when they range across varied “knowledge pools” of disciplinary expertise (Reagans & McEvily, 2003), span organizational boundaries (Tortoriello & Krackhardt, 2010), connect diverse genre styles (De Vaan et al., 2015), or involve different technological domains (Aral & Van Alstyne, 2011; Ter Wal, Alexy, Block, & Sandner, 2016).

In shifting from one-mode networks of collaboration to two-mode networks of competition, we were able to investigate afresh what combination of dyadic and triadic structural patterns allows actors to cope more effectively with uncertainty.

Limitations and future research

Our work builds on several premises. First, we assume that attention among pairs of competitors depends primarily on their structural competition patterns. Specifically, we expected that within a dyad, two competitors would pay equal attention to each other. Similarly, we assume that dyads with similar competition intensity will be comparable in their mutual attention. Recent research has shown, however, that this is not necessarily always the case. Besides structural aspects of competition, subjective perspectives can influence rivalry. For example, holding constant the objective structure of competition, dyads of competitors can differ in their relational rivalry (Kilduff, 2019; Kilduff, Elfenbein, & Staw, 2010), competitive tension (Chen et al., 2007), and identity domains (Livengood & Reger, 2010). Based on these factors, their mutual attention might vary.

Moreover, attention can be unevenly distributed within the dyad. For example, ego and alter might perceive differently the degree to which they compete (Thatchenkery & Katila, 2021). In our work, we highlighted a third unexplored difference. Attention among rivals is contingent on market uncertainty. Thus, the same dyad of competitors can vary in their mutual attention depending on the market on which they are focusing. Future research can integrate our market-centric approach and combine it with the established differences between and within dyads.

Building on our attention-based theory, future research can explore the role of status differences. On the one hand, high-status actors might be less sensitive to market uncertainty and less prone to paying attention to other interpretations (Bothner, Kim, & Smith, 2012). On the other hand, high-status actors might leverage the superior resources carried by their status to tap into the benefits of their network position (Prato & Ferraro, 2018).

Future research can also explore the effect of the broader observational network in which a competitor is embedded. In our study, we focused on the benefits of observing immediate rivals in a specific market. Nevertheless, the rivals of one’s rivals can also affect how a focal actor sees that market. Future research can explore the effect of being at the core or the periphery of the observational network or how other measures such as network centrality affect accuracy in market interpretations.

Finally, we assume that the relationship between an analyst’s observational network and greater accuracy in assessing a market’s prospects is the outcome of observational learning. Ethnographic research in trading rooms has shown that this is a reasonable assumption. In the study of the derivatives trading room of a major international investment bank on Wall Street, Beunza and Stark (2012) showed how traders learn from each other through a process of “reflexive modelling.” Similarly, Callon and Muniesa (2005) have proposed that markets are calculative collective devices and Knorr Cetina and Bruegger (2002) contend that they are constituted of an epistemic community observing and making sense of each other’s interpretations. Nonetheless, future research can explore observational learning among analysts more closely and analyze the text of analysts’ reports with natural language processing techniques such as topic modeling. For example, our theory would suggest that the topics in analysts’ reports who compete in several stocks will become more similar over time as they learn more from each other, and the analysts’ who enjoy exposure to non-redundant competitors would produce reports covering a greater number and/or more diverse topics.

Conclusion

Our theorization of network structures as scopic devices contributes to the observational perspective in the social studies of finance (Beunza & Stark, 2004; Knorr Cetina, 2003; Knorr Cetina & Bruegger, 2002; MacKenzie, 2011) by inviting researchers to reflect upon the specific role of network configurations built around observation patterns. In engaging in such an effort and in studying how specific structures can offer a privileged viewpoint to actors, we emphasize that not all network scopes are the same: As with microscopes, telescopes, and periscopes, network structures will differ in their properties and in the kinds of solutions or obstacles they pose.

In developing the conceptual and methodological tools for such a scopic network analysis, we focused on two structures—dyadic and triadic closure—each of which operates as a kind of optic. Consistent with our metaphor, each of these concepts refers to a notion of exposure, specifically to different patterns of exposure to the views or behavior of one’s competitors. Dyadic closure functions as a lens that regulates depth of focus: by increasing exposure to competitors’ behavior in different domains, dyadic closure sharpens vision of the competitors’ tacit knowledge. In contrast, triadic closure modulates peripheral vision: by increasing reciprocal exposure among one’s competitors, triadic closure regulates the diversity of behaviors to which a focal actor is exposed.

It is important to emphasize that we did not argue that the distinctive combination of closed dyads and open triads would be equally performance-enhancing across all circumstances. Instead, we explicitly specified that beneficial constraints for actors in the closed dyads/open triads structures should be most advantageous in more challenging, high risk-high reward, volatile situations. Indeed, our analysis confirms that analysts in the multiple exposure/diverse exposure network locations produced predictions yielding abnormally high returns.

Expressed in scopic terms, the optics that allow depth of focus and enhanced peripheral vision offered by high dyadic closure and low triadic closure enable actors to see through the fog of uncertainty. Like the optics of the human eye, which allowed our ancestral hunters and gatherers to have focused attention that was simultaneously alert to peripheral movement, the properties of scopic networks can yield considerable benefits. We demonstrated that actors embedded in such a network structure are in a privileged location to interpret information in turbulent situations quickly. Poised to find patterns when others hear noise, they can make use of dissonant messages to deal with uncertainty more effectively.

Footnotes

Acknowledgements

Our thanks to Byungkyu Lee, Josh Whitford, and other participants in the CODES seminar (Department of Sociology, Columbia University) and to Shahnawaz Akhtar for his work as research assistant. We also thank the editor in chief Renate Meyer, the associate editor Ha Hoang, and the three anonymous reviewers for their feedback and guidance during the review process.

Author Note

The authors benefited from comments received on prior versions of this manuscript in numerous conferences such as the annual meetings of the American Sociological Association August 2011 and August 2019; the 6th Annual Conference on Economic Sociology, Chapel Hill, October 2018; the Academy of Management Meeting 2013; the Junior Faculty Organization Theory Conference, Carnegie Mellon University 2016; and the 2nd Organization Theory Conference, USI Lugano, 2015 as well as seminars at HEC, IESE, ESMT and ESADE Business School.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article. Research for this paper was supported by an Advanced Research Grant from the European Research Council, BLINDSPOT project, grant number 695256.

Notes

Author biographies

Matteo Prato is associate professor at ESADE Business School. His current research interests lie at the intersection of economic sociology and organization theory, with particular emphasis on the role that social structures (e.g., status hierarchies, social networks, and categorization systems) play in shaping market valuations and in enhancing actors’ performance.

David Stark is Arthur Lehman Professor of Sociology at Columbia where he directs the Center on Organizational Innovation. He is also professor of social science at the University of Warwick where he recently directed a five-year project on “Diversity and Performance: Networks of Cognition in Markets and Teams” supported by a grant from the European Research Council. Stark’s recent publications include The Performance Complex: Competition and Competitions in Social Life (Oxford, 2020); “Racial Attention Deficit” (with Sheen Levine and Charlotte Reypens in Science Advances, 2021); “Algorithmic Management in the Platform Economy” (with Ivana Pais in Sociologica, 2021); “Moments of Identity: The Dynamics of Artist, Persona, and Audience in Electronic Music” (with Giovanni Formilan in Theory & Society, 2021); and “Testing and Being Tested in Pandemic Times” (Sociologica, 2020).

References

Ahuja

Gautam

(2000). Collaboration networks, structural holes, and innovation: A longitudinal study. Administrative Science Quarterly, 45, 425–455.

Aral

Sinan

Van Alstyne

Marshall

(2011). The diversity-bandwidth trade-off. American Journal of Sociology, 117, 90–171.

Bandura

Albert

(1977). Social learning theory. Englewood Cliffs, NJ: Prentice-Hall.

Baum

Joel A. C.

Korn

Helaine J.

(1996). Competitive dynamics of interfirm rivalry. Academy of Management Journal, 39, 255–291.

Baum

Joel A. C.

Korn

Helaine J.

(1999). Dynamics of dyadic competitive interaction. Strategic Management Journal, 20, 251–278.

Bellego

Christophe

Benatia

David

Pape

Louis D.

(2021). Dealing with logs and zeros in regression models. CREST-Série des Documents de Travail (2019-13).

Bernheim

B. Douglas

Whinston

Michael D.

(1990). Multimarket contact and collusive behavior. The RAND Journal of Economics, 21, 1–26.

Beunza

Daniel

Stark

David

(2004). Tools of the trade: The socio-technology of arbitrage in a Wall Street trading room. Industrial and Corporate Change, 13(1), 369–401.

Beunza

Daniel

Stark

David

(2012). From dissonance to resonance: Cognitive interdependence in quantitative finance. Economy and Society, 41, 383–417.

10.

Boehmke

Frederick J.

Chyzh

Olga

Thies

Cameron G.

(2016). Addressing endogeneity in actor-specific network measures. Political Science Research and Methods, 4, 123–149.

11.

Boeker

Warren

Goodstein

Jerry

Stephan

John

Murmann

Johann P

. (1997). Competition in a multimarket environment: The case of market exit. Organization Science, 8, 126–142.

12.

Boni

Leslie

Womack

Kent L.

(2006). Analysts, industries, and price momentum. Journal of Financial and Quantitative Analysis, 41, 85–109.

13.

Bothner

Matthew S.

Kim

Young K.

Smith

Edward B.

(2012). How does status affect performance? Status as an asset vs. status as a liability in the PGA and NASCAR. Organization Science, 23, 416–433.

14.

Botelho

Tristan L.

(2018). Here’s an opportunity: Knowledge sharing among competitors as a response to buy-in uncertainty. Organization Science, 29, 1033–1055.

15.

Bowers

Anne H.

Greve

Henrich R.

Mitsuhashi

Hitoshi

Baum

Joel A. C.

(2014). Competitive parity, status disparity, and mutual forbearance: Securities analysts’ competition for investor attention. Academy of Management Journal, 57, 38–62.

16.

Burt

Ronald S.

(1992). Structural holes: The social structure of competition. Cambridge, MA: Harvard University Press.

17.

Burt

Ronald S.

(2004). Structural holes and good ideas. American Journal of Sociology, 110, 349–399.

18.

Burt

Ronald S.

(2010). Neighbor networks: Competitive advantage local and personal. New York: Oxford University Press.

19.

Callon

Michel

Muniesa

Fabian

(2005). Peripheral vision: Economic markets as calculative collective devices. Organization Studies, 26, 1229–1250.

20.

Chen

Ming-Jer

Kuo-Hsien

Tsai

Wenpin

(2007). Competitive tension: The awareness-motivation-capability perspective. Academy of Management Journal, 50, 101–118.

21.

Chen

Victor Z.

Musacchio

Aldo

Sali

(2019). A principals-principals perspective of hybrid leviathans: Cross-border acquisitions by state-owned MNEs. Journal of Management, 45, 2751–2778.

22.

Clement

Michael B.

Tse

Senyo Y.

(2005). Financial analyst characteristics and herding behavior in forecasting. Journal of Finance, 60, 307–341.

23.

Correia

Sergio

Guimarães

Paulo

Zylkin

Tom

(2020). Fast Poisson estimation with high-dimensional fixed effects. Stata Journal, 20, 95–115.

24.

De Vaan

Mathijs

Stark

David

Vedres

Balazs

(2015). Game changer: The topology of creativity. American Journal of Sociology, 120, 1144–1194.

25.

Denrell

Jerker

(2003). Vicarious learning, undersampling of failure, and the myths of management. Organization Science, 14, 227–243.

26.

Dobusch

Laura

Dobusch

Leonhard

Müller-Seitz

Gordan

(2019). Closing for the benefit of openness? The case of Wikimedia’s open strategy process. Organization Studies, 40, 343–370.

27.

Edwards

Corwin D.

(1955). Conglomerate bigness as a source of power. In Stigler

George J.

(Ed.), Business concentration and price policy (pp. 331–360). Princeton, NJ: Princeton University Press.

28.

Friedkin

Noah E.

Johnsen

Eugene C.

(2011). Social influence network theory: A sociological examination of small group dynamics. New York: Cambridge University Press.

29.

Gavetti

Giovanni

Levinthal

Daniel A.

Rivkin

Jan W.

(2005). Strategy making in novel and complex worlds: The power of analogy. Strategic Management Journal, 26, 691–712.

30.

Gavetti

Giovanni

Warglien

Massimo

(2015). A model of collective interpretation. Organization Science, 26, 1263–1283.

31.

Gentner

Dedre

(2010). Bootstrapping the mind: Analogical processes and symbol systems. Cognitive Science, 34, 752–775.

32.

Gimeno

Javier

(1999). Reciprocal threats in multimarket rivalry: Staking out ‘spheres of influence’ in the US airline industry. Strategic Management Journal, 20, 101–128.

33.

Gimeno

Javier

Woo

Carolyn Y.

(1996). Hypercompetition in a multimarket environment: The role of strategic similarity and multimarket contact in competitive de-escalation. Organization Science, 7, 322–341.

34.

Goodman

Noah D.

Ullman

Tomer D.

Tenenbaum

Joshua B.

(2011). Learning a theory of causality. Psychological Review, 118, 110–119.

35.

Goyal

Amit

Santa-Clara

Pedro

(2003). Idiosyncratic risk matters! Journal of Finance, 58, 975–1007.

36.

Granovetter

Mark S.

(1973). The strength of weak ties. American Journal of Sociology, 78, 1360–1380.

37.

Greve

Henrich R.

(2005). Interorganizational learning and heterogeneous social structure. Organization Studies, 26, 1025–1047.

38.

Greve

Henrich R.

Taylor

Alva

(2000). Innovations as catalysts for organizational change: Shifts in organizational cognition and search. Administrative Science Quarterly, 45, 54–80.

39.

Hansen

Morten T.

(1999). The search-transfer problem: The role of weak ties in sharing knowledge across organization subunits. Administrative Science Quarterly, 44, 82–111.

40.

Vivianna F.

von Krogh

Georg F.

Sirén

Charlotta

(2022). Expertise diversity, informal leadership hierarchy, and team knowledge creation: A study of pharmaceutical research collaborations. Organization Studies, 43, 907–930.

41.

Hoberg

Gerard

Phillips

Gordon

(2010). Product market synergies and competition in mergers and acquisitions: A text-based analysis. Review of Financial Studies, 23, 3773–3811.

42.

Hong

Harrison

Kacperczyk

Marcin

(2010). Competition and bias. Quarterly Journal of Economics, 125, 1683–1725.

43.

Huygens

Marc

Van Den Bosch

Frans A. J.

Volberda

Henk W.

Baden-Fuller

Charles

(2001). Co-evolution of firm capabilities and industry competition: Investigating the music industry, 1877–1997. Organization Studies, 22, 971–1011.

44.

Jayachandran

Satish

Gimeno

Javier

Varadarajan

P. Rajan

(1999). The theory of multimarket competition: A synthesis and implications for marketing strategy. Journal of Marketing, 63(3), 49–66.

45.

Kaplan

Sarah

(2011). Research in cognition and strategy: Reflections on two decades of progress and a look to the future. Journal of Management Studies, 48, 665–695.

46.

Kilduff

Gavin J.

(2019). Interfirm relational rivalry: Implications for competitive strategy. Academy of Management Review, 44, 775–799.

47.

Kilduff

Gavin J.

Elfenbein

Hillary A.

Staw

Barry M.

(2010). The psychology of rivalry: A relationally dependent analysis of competition. Academy of Management Journal, 53, 943–969.

48.

Knorr Cetina

Karin

(2003). From pipes to scopes: The flow architecture of financial markets. Distinktion: Scandinavian Journal of Social Theory, 4(2), 7–23.

49.

Knorr Cetina

Karin

Bruegger

Urs

(2002). Global microstructures: The virtual societies of financial markets. American Journal of Sociology, 107, 905–950.

50.

Levinthal

David A.

March

James G.

(1993). The myopia of learning. Strategic Management Journal, 14(S2), 95–112.

51.

Livengood

R. Scott

Reger

Rhonda K.

(2010). That’s our turf! Identity domains and competitive dynamics. Academy of Management Review, 35, 48–66.

52.

MacKenzie

Donald

(2011). The credit crisis as a problem in the sociology of knowledge. American Journal of Sociology, 116, 1778–1841.

53.

Meltzoff

Andrew N.

Waismeyer

Anna

Gopnik

Alison

(2012). Learning about causes from people: Observational causal learning in 24-month-old infants. Developmental Psychology, 48, 1215–1228.

54.

Mikhail

Michael B.

Walther

Beverly R.

Willis

Richard H.

(1997). Do security analysts improve their performance with experience? Journal of Accounting Research, 35, 131–157.

55.

Nelson

Richard R.

Winter

Sidney G.

(1982). The Schumpeterian trade-off revisited. American Economic Review, 72, 114–132.

56.

Ozmel

Umit

Yavuz

Deniz

Trombley

Tim

Gulati

Ranjay

(2020). Interfirm ties between ventures and limited partners of venture capital funds: Performance effects in financial markets. Organization Science, 31, 698–719.

57.

Payne

Jeff L.

Thomas

Wayne B.

(2003). The implications of using stock-split adjusted I/B/E/S data in empirical research. Accounting Review, 78, 1049–1067.

58.

Phillips

Damon J.

Zuckerman

Ezra W.

(2001). Middle-status conformity: Theoretical restatement and empirical demonstration in two markets. American Journal of Sociology, 107, 379–429.

59.

Podolny

Joel M.

(1993). A status-based model of market competition. American Journal of Sociology, 98, 829–872.

60.

Podolny

Joel M.

(2001). Networks as the pipes and prisms of the market. American Journal of Sociology, 107, 33–60.

61.

Podolny

Joel M.

(2008). Resurrecting images from the past? Comment on Reagans and Zuckerman. Industrial and Corporate Change, 17, 971–977.

62.

Polanyi

Michael

(1966). The logic of tacit inference. Philosophy, 41(155), 1–18.

63.

Porac

Joseph F.

Thomas

Howard

Baden-Fuller

Charles

(1989). Competitive groups as cognitive communities: The case of Scottish knitwear manufacturers. Journal of Management Studies, 26, 397–416.

64.

Prato

Matteo

Ferraro

Fabrizio

(2018). Starstruck: How hiring high-status employees affects incumbents’ performance. Organization Science, 29, 755–774.

65.

Reagans

Ray

McEvily

Bill

(2003). Network structure and knowledge transfer: The effects of cohesion and range. Administrative Science Quarterly, 48, 240–267.

66.

Reagans

Ray

Zuckerman

Ezra W.

(2001). Networks, diversity, and productivity: The social capital of corporate R&D teams. Organization Science, 12, 502–517.

67.

Reed

Richard

DeFillippi

Robert J.

(1990). Causal ambiguity, barriers to imitation, and sustainable competitive advantage. Academy of Management Review, 15, 88–102.

68.

Silva

Joao S.

Tenreyro

Silvana

(2006). The log of gravity. Review of Economics and Statistics, 88, 641–658.

69.

Sorenson

Olav

(2003). Interdependence and adaptability: Organizational learning and the long-term effect of integration. Management Science, 49, 446–463.

70.

Stuart

Toby E.

Hoang

Hybels

Ralph C.

(1999). Interorganizational endorsements and the performance of entrepreneurial ventures. Administrative Science Quarterly, 44, 315–349.

71.

Ter Wal

Anne L.

Alexy

Oliver

Block

Jorn

Sandner

Philipp G.

(2016). The best of both worlds: The benefits of open-specialized and closed-diverse syndication networks for new ventures’ success. Administrative Science Quarterly, 61, 393–432.

72.

Thatchenkery

Sruthi

Katila

Riitta

(2021). Seeing what others miss: A competition network lens on product innovation. Organization Science, 32, 1149–1390.

73.

Tortoriello

Marco

Krackhardt

David

(2010). Activating cross-boundary knowledge: The role of Simmelian ties in the generation of innovations. Academy of Management Journal, 53, 167–181.

74.

Tripsas

Mary

Gavetti

Giovanni

(2000). Capabilities, cognition, and inertia: Evidence from digital imaging. Strategic Management Journal, 21, 1147–1161.

75.

Tsai

Wenpin

Kuo H.

Chen

Ming J.

(2011). Seeing through the eyes of a rival: Competitor acumen based on rival-centric perceptions. Academy of Management Journal, 54, 761–778.

76.

Uzzi

Brian

(1996). The sources and consequences of embeddedness for the economic performance of organizations: The network effect. American Sociological Review, 61, 674–698.

77.

Vedres

Balazs

(2017). Forbidden triads and creative success in jazz: The Miles Davis factor. Applied Network Science, 2(1), 31.

78.

Walker

Caren M.

Gopnik

Alison

(2014). Toddlers infer higher-order relational principles in causal learning. Psychological Science, 25, 161–169.

79.

White

Harrison C.

(1981). Where do markets come from? American Journal of Sociology, 87, 517–547.

80.

Fang F.

(2008). Analyst coverage and earnings management. Journal of Financial Economics, 88, 245–271.

81.

Zuckerman

Ezra W.

(1999). The categorical imperative: Securities analysts and the illegitimacy discount. American Journal of Sociology, 104, 1398–1438.

82.

Zuckerman

Ezra W.

(2004). Structural incoherence and stock market activity. American Sociological Review, 69, 405–432.