An Interaction Ritual Theory of Social Resource Exchange: Evidence from a Silicon Valley Accelerator

The Social Dilemma in Generalized Exchange

Social exchange theory conceptualizes social life as a series of exchanges that are structured in distinct ways (Emerson, 1972). The theory is among the most widely used conceptual lenses in sociology and management scholarship to understand exchange relationships among actors (e.g., Gulati and Sytch, 2007; Willer, Flynn, and Zak, 2012). Two specific exchange structures identified in social exchange theory are direct and generalized exchange. Direct exchange involves a transfer of resources between two actors with the expectation of bilateral reciprocity (Ekeh, 1974). Direct exchange structures can be negotiated—referring to a simultaneous exchange between two actors after bargaining over terms—or reciprocal—referring to sequential acts of giving and receiving within a dyad (Molm and Cook, 1995).

Generalized exchange, in contrast, involves indirect giving and receiving of resources among actors who belong to a group (i.e., A’s unilateral giving to B is reciprocated not by B but by another member of the group) (Ekeh, 1974). Generalized exchange can take several forms. In chain generalized exchange, resource exchange flows in one direction to complete a circular pattern among a group of actors (Bearman, 1997; Schilke and Rossman, 2018). In net generalized exchange, actors take turns contributing to group objectives (Cheshire, 2007). Finally, pure generalized exchange does not follow any specific pattern (Willer, Flynn, and Zak, 2012); at the time of giving, it is not apparent to the giver which of the group members might reciprocate the favor, in what form, or when (Takahashi, 2000).

Accelerators particularly encourage pure generalized exchange among entrepreneurs from a cohort who undergo similar challenges (Cohen, Bingham, and Hallen, 2019) and are encouraged to “help its other peer organizations build their ventures” (Amezcua et al., 2013: 1636). Through their objectives and mission statements, accelerators implicitly or explicitly signal to entrepreneurs that they expect them to give “first to others . . . with no specific expectations of return” (see footnote 2; https://www.techstars.com/). In the accelerators studied by Cohen, Bingham, and Hallen (2019: 834), founders refer to the generalized exchange promoted by accelerators as “(t)he vibe [to] ‘help everyone at all cost’.” Moreover, accelerators suggest that unilateral exchanges may be indirectly reciprocated by other peers, drawing entrepreneurs’ attention to a potentially supportive cohort of peers who will become the foundation of their support system (Amezcua et al., 2013; Hallen, Cohen, and Bingham, 2020). ³ As is typical of pure generalized exchanges, when and in what form any of their peers might reciprocate is not explicitly specified (Lévi-Strauss, 1969; Ekeh, 1974; Yamagashi and Cook, 1993). While an entrepreneur who gives to a peer may, in the future, receive from someone else in the cohort, there is no guarantee he or she will.

This uncertainty exposes generalized exchange structures to a social dilemma: in addition to the risk of nonreciprocity, participants have to overcome the temptation to free-ride (Yamagashi and Cook, 1993; Takahashi, 2000). On the one hand, rational self-interested actors may succumb to the temptation to free-ride, as they will be better off if they appropriate all the benefits of the system without giving to others themselves. On the other hand, participants who perceive the risk of nonreciprocity in the system to be high will refrain from engaging in unilateral exchanges. Recent research suggests that this social dilemma is very real in resource exchange networks in accelerators (Grimes, 2018; Cohen, Bingham, and Hallen, 2019). In their study examining how accelerator designs mitigate the bounded rationality of entrepreneurs, Cohen, Bingham, and Hallen (2019) suggested that although entrepreneurs learned from their peers, they also felt pressure to outperform them. By providing the opportunity to observe each other’s progress in open workspaces, accelerators facilitate information sharing and learning from peers. But exposing entrepreneurs to the progress of peers also creates peer pressure to outperform them. This competitive component is particularly salient, as entrepreneurs in accelerators may imitate each other’s best practices (Cohen, Bingham, and Hallen, 2019) or pivot into their peers’ product domains (Grimes, 2018). While entrepreneurs stand to gain from sharing social resources, they may also incur costs by losing their competitive advantage in the process.

Given these concerns, not only have scholars expressed surprise that generalized exchange systems are established in the first place, but they are equally puzzled by the stability of some generalized exchanges (Takahashi, 2000). The proponents of generalized exchange trace this structure’s potential stability back to the affective orientation inherent in the exchanges involved (Lévi-Strauss, 1969; Ekeh, 1974; Molm, Collett, and Schaefer, 2007). Considering the nontrivial risk of nonreciprocity, the very act of unilateral giving in generalized exchanges may convey an affective orientation toward the recipient above and beyond the instrumental value attached to the resource. Similarly, recipients tend to treat the benefits they receive from other participants as unilateral gifts, resulting in attachment to the group. Accordingly, generalized exchange can produce a stable micro social order comprising a recurrent pattern of exchange based on individual actors’ sense of belonging to the group, their enduring relationships with each other, and their positive feelings about the exchanges (Molm, Collett, and Schaefer, 2007; Lawler, Thye, and Yoon, 2008).

Social exchange scholarship, however, has produced inconclusive empirical findings on the stability of generalized exchanges. For instance, whereas Lawler, Thye, and Yoon (2000) found that generalized exchanges are considerably less stable than direct exchanges, Molm, Collett, and Schaefer (2007) produced evidence to the contrary. Interestingly, even though Molm, Collett, and Schaefer (2007) found evidence for the persistence of generalized exchanges, there was substantial variation. While unilateral giving crumbled quickly in some generalized exchange systems, it was remarkably stable in others. If the affective orientation communicated by the act of giving was sufficient to generate a stable micro social order, the instability of some of the generalized exchange systems is difficult to explain. Resolving this discrepancy has not been easy, as much extant social exchange research has exogenously assigned actors to specific exchange structures in lab experiments, as though social exchanges are independent of the situations and the context in which the actors are embedded. We depart from this practice by closely examining the evolution of situational dynamics surrounding acts of giving and the nature of relationships among exchange participants. As Baldassarri (2015: 363) put it, “if one wants to understand the mechanisms through which cooperation emerges in a specific social setting, it is important to focus on how patterns of social relations affect actors’ motives . . . or their expectations about others’ behavior.”

To gain a deeper conceptual understanding of the emergence and evolution of generalized exchanges, we turn to interaction ritual theory, which uses as its starting point the dynamics of social situations to explain the emergence of stable social order. By examining emergent social situations that pave the way for future interactions and social ties, interaction ritual theory helps elucidate the founding conditions and ensuing dynamics underlying stable vs. unstable generalized exchanges.

Interaction Rituals and Social Order

At its core, interaction ritual theory explains how social encounters, ranging from small to very large gatherings, create shared experiences, common fate, and ultimately strong identification with the group (Durkheim, 1912; Goffman, 1967; Collins, 2004). By providing shared emotional experiences to participants, interaction rituals make membership in a group meaningful. Interaction ritual theory originates in Durkheim’s (1912) work on religious rituals. By allowing actors to interact on the basis of intensely shared emotions that endure long after the event that triggered them, religious rituals promote identification with the group and moral obligation toward fellow group members. Goffman (1967) extended Durkheim’s (1912) work to the secular domain of everyday micro-social encounters, including smaller gatherings, such as informal parties, and dyadic social encounters, such as dinners.

Collins (2004) synthesized Durkheim’s and Goffman’s ideas into interaction ritual theory. According to this theory, interaction rituals are activated when actors co-located in a physical space focus their attention on a common activity and experience shared emotions (e.g., joy, amusement, solemnity). While common focus refers to the focus of co-located actors’ attention on a particular activity or object that they become mutually aware of as they observe each other doing the same thing, shared emotions refers to an emotional experience that co-located actors share when they direct their collective attention to a certain activity or object (Collins, 2004). Although intensely shared emotions experienced in social events may carry over to day-to-day interactions, Collins (2004) noted that collective sentiments will be reduced to mere memories if they are not renewed in subsequent rituals. While some interaction rituals can persist routinely over time in a predetermined fashion (e.g., routinely held formal meetings), other interaction rituals (e.g., dinners, informal meetings) need not recur regularly. Recurrence of interaction rituals over time, whether predetermined or not, constitutes an interaction ritual chain (Collins, 2004).

As individuals move from one social encounter to the next, they derive high emotional energy from rituals that provide a sense of belonging to the group and low emotional energy from rituals that make them feel excluded from the group (Collins, 2004). Goffman (1959, 1967) shed light on how individuals respond to the emotional energy derived from social encounters by examining the finer details in a sequence of micro social encounters, ranging from dyadic to small-group interactions. Goffman (1967) argued that in every micro interaction, individuals project a version of themselves. For instance, “to ask for a job, a loan of money, or a hand in marriage is to project an image of self as worthy” (Goffman, 1967: 106). When individuals find that their projected self is discredited in a social encounter, they feel shame and embarrassment, while acceptance of their projected self generates pride. Goffman (1967: 44) pointed out that “refusal of a request for assistance becomes a delicate matter, potentially destructive of the asker’s face.” Individuals react to shame by resorting to anger in subsequent social encounters or even avoiding the interaction altogether (Scheff, 1990).

Overall, research on interaction rituals shows that a common focus and shared emotions can generate strong identification with the group and that individuals are drawn to those social encounters from which they derive high emotional energy (see Weiniger and Lizardo, 2019, for a review). While prior research has demonstrated how interaction rituals enable cooperation among individuals, it has often assumed that the costs of cooperation are trivial (e.g., Furnari, 2014). Most prior organizational research that has used interaction ritual theory to understand cooperation in groups has been conducted in unproblematic contexts involving interdependent individuals who work toward a common goal and has mainly focused on positive emotions (e.g., Metiu and Rothbard, 2013). Yet in mixed-motive contexts, such as accelerators, cooperation carries substantial costs and risks to individual actors. We present evidence from a Silicon Valley accelerator and build an interaction ritual theory of the emergence of generalized exchange in such mixed-motive contexts.

Research Setting

Accelerators have become an increasingly important phenomenon in the start-up ecosystem worldwide (Hallen, Cohen, and Bingham, 2020). A typical accelerator runs start-up camps for a finite period and brings together, through a competitive selection process, entrepreneurs who are unlikely to have known each other prior to entering the accelerator (Cohen, Bingham, and Hallen, 2019). As elaborated above, the interests of entrepreneurs both align and conflict in the accelerator setting, making it representative of mixed-motive contexts. From September 2014 to April 2015, the lead author conducted ethnographic fieldwork in a well-known Silicon Valley accelerator, in which entrepreneurs obtained access to mentors, investors, and corporate clients. All participants in our study were guaranteed confidentiality and anonymity. The three mentors were in-house experts with substantial experience in accelerating start-ups. The accelerator co-located entrepreneurs in an open workspace in which they encountered investors and corporate clients either as guest speakers in weekly meetings or through mentor referrals. At the time of our study, the accelerator focused on early-stage technology-related start-ups seeking seed funding. Each quarter, the accelerator organizes start-up camps for nascent entrepreneurs typically lasting ten weeks and ending with a demo day. The accelerator offers two different types of start-up camps—one in which entrepreneurs obtain equity investment, and another in which start-ups incorporated outside the U.S. are required to pay a fee to be part of the Silicon Valley experience.

Sampling Design

The spring 2015 cohort, the focus of our investigation, included three distinct start-up camps: Camp A, entrepreneurs staying on from the fall 2014 cohort who obtained equity investment from the accelerator (the accelerator allowed entrepreneurs to remain at the accelerator for up to four months after demo day); Camp B, a new batch of entrepreneurs incorporated outside the U.S. who paid a fee to be part of the accelerator; and Camp C, a new batch of entrepreneurs who obtained equity investment from the accelerator. Table 1 provides details on the features and composition of these three camps.

OPEN IN VIEWER

Table 1.

Descriptives by Camp

	Camp A	Camp B	Camp C
Average number of cofounders	2.00	2.27	2.69
Percentage of CEOs	54%	50%	35%
Percentage of CTOs	18%	25%	35%
Percentage of males	91%	83%	85%
Average start-up age	1.77 years	1.75 years	1.25 years
% of start-ups with some investor funding	27%	16%	15%
% of entrepreneurs who are white	55%	67%	50%
% of entrepreneurs whose country of origin is U.S.	27%	17%	46%
Equity from accelerator	Yes	No	Yes
Structured weekly formal meetings	No*	No	Yes
% of start-ups that survived 2 years post demo day	71%	87%	30%
Business
Photo sharing			18%
Fintech	43%		18%
Internet of things			9%
Crowdfunding			9%
Travel	14%	29%	9%
Education			9%
Professional services		29%	9%
Social media			18%
Retail	14%	29%
Big data	14%
Legal	14%
Voice recognition		14%
Number of founders	11	12	26

*

Although Camp A attended the mandatory weekly formal meetings in the fall of 2014, they were not attending any weekly formal meetings during the quantitative data collection period in the spring of 2015.

To check the robustness of our findings and clarify mechanisms, we also consider the fall 2014 cohort—Camps X, Y, and AF—for which we had data from preliminary interviews and observations. ⁴ We use the label “Camp AF” because the entrepreneurs from this fall camp who remained at the accelerator became Camp A in spring 2015. The program structures of Camps X and AF were comparable to those of Camp C, with entrepreneurs obtaining equity investment from the accelerator; the program structure of Camp Y was comparable to that of Camp B, with entrepreneurs being incorporated outside the U.S.

Alternative design choices might have involved a study of multiple accelerators with an experimental intervention in some of them or an experiment involving an intervention event in the middle of an accelerator’s training efforts. But studying a realistic case in a natural setting aligned better with our objective to observe the organic evolution of exchange structures over time. In particular, this approach allowed us to pursue our broad, open-ended research questions while creating an intimate connection with and understanding of the empirical phenomenon of our study in situ (Suddaby, Bruton, and Si, 2015; Huang, 2018). The Silicon Valley accelerator we studied allowed us to observe in a natural setting how the emergent social situations within social exchange units shaped the nature of exchange relationships among co-located peers, culminating in the emergence or nonemergence of generalized exchange. We investigated these emergent processes over time in one cohort and checked for robustness of our qualitative findings with comparable data from a previous cohort. We then deductively examined the emergent theory using quantitative analysis.

Data Sources

We used several data sources that complemented each other or allowed for triangulation, including interviews and observations.

Abductive Analysis

We used an abductive approach to interpret our ethnographic data and construct our theoretical model. Abductive analysis was particularly suitable in light of the inconclusiveness of existing social exchange scholarship, limited evidence on social resource exchange among peers in the accelerator setting, and emerging insights from our data that could not be explained by existing research (Kreager et al., 2017; Shepherd and Suddaby, 2017). Abduction refers to an iterative inferential process that spurs both inductive and deductive reasoning (Suddaby, 2006), such that inductive grounded open coding generates theoretical leads by identifying unexpected themes that diverge from existing theory, which are pursued deductively through a re-analysis of existing data or through new data-gathering rounds (Timmermans and Tavory, 2012).

Our abductive analysis used a sequential mixed-methods strategy, proceeding from qualitative to quantitative analysis (e.g., Huang and Pearce, 2015; Kreager et al., 2017). We began our qualitative analysis with open coding of interviews and field observations to form first-order codes, which involved systematically categorizing statements that were similar and dissimilar across Camps A, B, and C (Pratt, 2009). Our early iterations involved looking for statements describing social resource exchange among peers. For instance, statements from entrepreneurs across all camps indicated active resource seeking, e.g., “needed some information.” Yet from the way entrepreneurs completed their resource request statements, it was clear that their peers’ responses were starkly different across camps, e.g., “and got some solid input” (Camp B) vs. “were not responsive” (Camp C). At this stage, we also coded words and phrases from our field notes that increased our understanding of the context and situations in which entrepreneurs’ relationships and social resource exchange experiences are embedded. Collective sentiments in different social events emerged as a dominant theme in our early coding process, resulting in fresh rounds of data collection pertaining to the nature of social events, e.g., informal events in Camps A and B. Over the course of revising our theorizing, we conducted new rounds of coding by revisiting subsets of the data.

To build our theoretical model, we employed temporal bracketing—”the decomposition of data into successive adjacent periods” (Langley, 1999: 703)—in view of the recurrence of interaction episodes in our data. Identifying such sequences of events was essential to creating structure in the theoretical model (Shepherd and Suddaby, 2017). Each interaction episode comprised distinct social events and subsequent workplace exchange dynamics associated with those events. This temporal decomposition of comparable episodes that recur over time allows replication of emergent theory (Langley, 1999). We were able to identify three distinct phases from the temporal ordering of our data. We treat the seven days following the new entrepreneur orientations as phase 0, as these early inter-camp interactions were similar across camps. In the weekend that followed, several entrepreneurs from Camps A and B engaged in informal activities, which marked the beginning of phase 1. The transition from phase 1 to phase 2 was mediated by virtuous (Camp A and B) and vicious (Camp C) exchange cycles set in motion by dyadic social encounters in the workspace. While a clear breakpoint in collective sentiments in a formal weekly meeting marked the beginning of phase 2 in Camp C, the transition to phase 2 in Camps A and B was more gradual.

We compared first-order codes within and across phases to identify common and distinct themes over time. To arrive at second-order codes from first-order codes, we iterated between themes and existing literature (Eisenhardt, 1989). For a literature that could speak to interaction dynamics spurred by collective sentiments in social events, we turned to interaction ritual theory (Durkheim, 1912; Goffman, 1967; Collins, 2004). Finally, we combined related second-order theoretical categories into aggregate dimensions. As an illustration of this process, consider the following statements from Camp C entrepreneurs: ⁶

We initially categorized statements 1 and 2 as indicative of first-order codes “tense about progress” and “unwilling to share information,” respectively, which tended to specifically apply to entrepreneurs in Camp C. Although both statements pertained to shared negative emotions, they were reflective of different phases in the program period. As we iterated between first-order themes and literature, these statements were included under a second-order code “shared emotions”—”tense” in phase 1 and “distrust” in phase 2—and finally under the aggregate dimension “tournament rituals.” Similar coding of statements pertaining to informal activities in Camps A and B resulted in the aggregate dimension “bonding rituals.” This coding process helped us build a theoretical model in which the exchange expectations set in meso interaction rituals shaped exchange dynamics in dyadic micro social encounters at the workspace, which subsequently spilled over into subsequent meso interaction rituals in a recursive process. This recursive process changed the nature of relationships between peers over time and culminated in distinct macro exchange structures. Moreover, we identified crossover paths by comparing peer relationships in Camp C with those in Camp AF from a previous cohort that had a similar program structure but different outcomes. Figure A1 in the Online Appendix illustrates our data structure. Table 2 contains representative first-order data underlying our theoretical categories.

OPEN IN VIEWER

Table 2.

Representative Quotes and Observational Entries Underlying Categories

Phase 0: Initial Situational Encounters
Comparable and distinct early interactions
Equal opportunity for interaction		Open office space, common areas for potential interaction, peers visible from any seating area. (Source: field notes)
Comparable workplace interactions		Similar inter-camp interaction patterns: Entrepreneurs sought to get to know their peers from other camps, e.g., interaction with peers from other camps who are seated proximally, responding to interactions initiated by other camp peers. (Source: field notes)
Distinct introductory formal events		Generating familiarity, Camp B:“I became familiar with what others were doing after the icebreaker game.” (Henry, Camp B)Highlighting strengths, Camp C: Emphasis on strength of start-ups. Building familiarity was not the main focus; participants did not even go through a basic round of introductions. (Source: field notes)
Phase 1: Emergence of Expectations and Dyadic Exchange Processes
Early bonding rituals		Early tournament rituals
Common focus: building familiarity	“We saw each other every day at work. Cafeterias were closed, and I guess we were both hungry, and it was almost weekend. We just decided to head out for dinner together. That’s when we hit it off.” (Sophia, Camp B)“We bumped into each other at one of those hacker house barbecues.” (Julian, Camp A)	Common focus: show of strength	Formal meeting segment 1: reporting on the past week’s progress, spelling out their goals for the subsequent week, feedback from the mentor and other start-ups, weekly peer rating. (Source: field notes)Formal meeting segment 2: pitching ideas to investors and clients, networking opportunities, establishing contacts, hearing advice and experiences from speakers. (Source: field notes)
Shared emotions: comfort	“We crashed at their place after the party. It was too late to go home.” (Simon, Camp B)“For a while I didn’t know what Ethan and Henry were doing. I could only discuss work after we got together a couple of times.” (Sophia, Camp B)	Shared emotions: tension, anticipation	Tension:“You come thinking you have found the right tagline. And then he destroys it. That was a lot of stress.” (Victor, Camp C)Anticipation:“Meeting new speakers was the best part. I looked forward to getting to know them.” (Harry, Camp C)
Dyadic exchange processes: bonding		Dyadic exchange processes: tournament
Expectation of unilateral exchange and fulfilled exchanges	“Once you spend time together, it is easier to reach out for support.” (Robert, Camp B)“I needed some information on the U.S. market and got some solid input from Benjamin.” (Henry, Camp B)	Expectation of negotiated exchange and failed exchanges	“We had very aggressive goals and didn’t have the time to help anyone. Besides, we are focused on building a product now and do not need help from anyone here.” (Harry, Camp C)“We hoped to sign up a lot of start-ups for our app. Most of them were not responsive.” (George, Camp C)
Phase 2. Exchange Expectation (Mis)Alignment and Micro Social Order
Subsequent bonding rituals		Subsequent tournament rituals
Common focus: show of camaraderie	“We were regulars at the AIR parties.” (Henry, Camp B)“We mostly hiked together. Oh, yeah. We did the hackathon together once. That was kind of fun.” (Mason, Camp A)	Common focus: show of strength	Formal meeting segment 1: reporting on the past week’s progress, spelling out goals for the subsequent week, feedback from the mentor and other start-ups, weekly peer rating. (Source: field notes)Formal meeting segment 2: pitching ideas to investors and clients, networking opportunities, establishing contacts, hearing advice and experiences from speakers. (Source: field notes)
Shared emotions: fellow feeling	“Finding money in the Valley is not as easy as it might seem. When others share their stories, it kind of nudges you to open up as well. Our journeys might be different, but the frustration is just the same.” (Tyler, Camp A)“We were all pretty excited to be at the hackathon.” (Cooper, Camp B)	Shared emotions: distrust, apathy	Distrust: In one session, when Tom asked Oliver about a platform used in their new video campaign, Oliver ended the exchange with a curt response: “So you are asking me about details of my campaign, which I am not sharing.” (Source: field notes)Apathy: Mentor announced one week, “How about [suggesting] a one-liner tagline for Investo on Facebook? We will vote for the best tagline next week.” The group’s Facebook page was silent; not a single entrepreneur suggested a tagline. (Source: field notes)
Bonding: stable micro social order		Tournament: unstable micro social order
Identification with group	“We are all in the same boat.” (Aaron, Camp A)“You hear about what a company has done. You have start-ups across the table that have made it big. Seeing them work inspires me.” (Simon, Camp A)	Disidentification from group	“We came to work every day simply because we didn’t have a table at home.” (Oleg, Camp C)“I feel we are strong in technology, and the funding landscape and needs of others are different. So I don’t think they are the right people for such advice.” (Charlie, Camp C)
Strong affective ties with peers	“We are like best friends here.” (Simon, Camp B)“Some of these guys have gone out of the way to help us. Coming in, I didn’t expect I’d feel this connection with folks here.” (Benjamin, Camp A)	Weak ties and detachment from peers	People stopped nodding or smiling at each other when they arrived at work. (Source: field notes)“I will invite you once or twice. But if you keep declining, I will think you are not interested and will stop inviting you.” (Ethan, Camp B, referring to Camp C founders)
Perception of resource gain	“Whenever I brainstorm with Hudson, I gain a lot of clarity.” (Parker from Camp B)“I have received inputs from Aaron, and they are usually good.” (Carter, Camp A)	Perception of resource drain	“I don’t believe that the Smartpak idea is worth pursuing, and even the investors feel that way. I have given them different ideas. But I do not know why they keep pursuing this track. I don’t think they realized that a lot of effort went into thinking about their problem. I advised them because I understood what the investor meant.” (Frank, Camp C)“I wanted to invest in Beat. I gave them some feedback. But they said they needed to focus on their product.” (Albert, Camp C)
Interaction Ritual Crossovers: Expectation Realignment
Consciously coordinated bonding rituals and expectation realignment		“In the beginning, they routed their interaction through me. They approached me when they needed help, and I connected them to others in the cohort. That’s how we became friendly, and I invited them to a baseball game and then to my birthday party.” (Jenny, intern Camp AF, fall 2014)“After Jenny invited us to the baseball game, I got together with others a few times. By demo day, I had buddies I could count on.” (Tyler, Camp AF)

In line with our abductive analysis, we deductively pursued the relationships that surfaced in our inductive theoretical model through quantitative analysis. Specifically, we tested whether interaction rituals systematically affected entrepreneurs’ perceptions of unilateral exchanges and whether workplace interactions mediated this relationship. Our quantitative findings supported the theoretical insights that emerged from our qualitative analysis while accounting for alternative explanations. Below, we discuss in detail our findings from our qualitative analysis followed by those from our quantitative analysis.

Phase 0. Initial Situational Encounters

Comparable early interactions across camps and asymmetric emergence of bonding rituals

Entrepreneurs from all three camps shared the same open workspace, which offered them ample opportunities to interact. Founders could see every table from virtually any location in the space, and they passed by other camps’ tables on the way to the watercooler—Figure 1 shows a seating map. For the first few days, entrepreneurs in all camps made comparable use of this opportunity for interaction; their first encounters with peers from other camps occurred in this space. As is typical of such interactions, seating proximity and homophily (i.e., focusing on peers from the same country or race) were key drivers of founders getting to know each other (Rivera, Soderstrom, and Uzzi, 2010).

OPEN IN VIEWER

Figure 1.

Seating Map

Despite these comparable initial levels of interaction, the emergence of informal activities across camps was asymmetric as the first weekend neared. Plans for informal weekend activities initiated by peers emerged quickly within and across Camps A and B but were conspicuously absent in Camp C. We discuss these differences more in the section on early bonding rituals in phase 1, but it was clear within just a few days that the familiarity established in the introductory formal event in Camp B made its founders more comfortable with each other than the founders in Camp C were. That level of comfort in Camp B also seemed to help those founders initiate and welcome informal events with peers from Camp A.

Camp C entrepreneurs did make earnest efforts to become acquainted with their peers during that first week. For instance, right after the new entrepreneur orientation, Harper volunteered to collect the Twitter handles, e-mail addresses, and LinkedIn profiles of all the founders in Camp C. Subsequently, he posted this information on the camp’s Facebook page. But Camp C’s entrepreneurs were not as familiar and comfortable with each other as Camp B’s were in that first week, and the result was that Camps A and B started engaging in informal events earlier than Camp C did.

Phase 1. Emergence of Expectations and Dyadic Exchange Processes

In this section, we explain how interactions within and between Camps A and B intensified while those even among peers within Camp C gradually dwindled. Figure 2 illustrates the difference in the weekly interaction trajectories of the three camps. In addition, we show how interactions within and between camps spurred further differences that ultimately affected the exchange of resources.

OPEN IN VIEWER

Figure 2.

Cumulative Workplace Interactions Initiated by Camp Members

Phase 2. Exchange Expectation (Mis)Alignment and Micro Social Order

Cycle of avoidance and shaming, non-emergence of generalized exchange, and dissolution of micro social order

In failed exchanges in Camp C, resource seekers and resource givers both felt rejected by their peers. Resource seekers felt rejected because peers declined their resource requests. Resource givers felt rejected because peers appeared to undervalue the resources they were offering—resources that often were unsolicited and/or did not match entrepreneurs’ needs. Accordingly, many resource givers found themselves in unilateral exchanges when they expected to be in bilateral reciprocal exchanges.

Failed exchanges became shaming rituals, as entrepreneurs who felt undervalued and rejected by their peers avoided their peers in order to stay away from such negative experiences (Goffman, 1959, 1967; Scheff, 1990; Collins, 2004). Entrepreneurs who were shamed severed ties with the peers responsible for the shaming and avoided others. In Figure 3, we compare the growth of cumulative workplace interactions initiated by resource seekers who experienced failed exchanges with those initiated by the rest of the entrepreneurs in Camp C. Although both categories had similar starting points, Camp C entrepreneurs who were rejected by their peers were quick to avoid further interactions with their peers in the open workspace. This contributed to a vicious cycle of shaming and avoidance. Their negative experiences at the workspace manifested as apathy and distrust in subsequent tournament rituals, culminating in the non-emergence of a generalized exchange system and an unstable micro social order. By the end of the study period, rapid dissolution of ties resulted in a sparsely connected network that was populated with more isolates than meaningful ties. Entrepreneurs disidentified with their peers, recalled weak ties with them, and described their unilateral giving as a “waste of time.”

OPEN IN VIEWER

Figure 3.

Cumulative Workplace Interactions Initiated by Camp C Entrepreneurs

Interaction Ritual Crossover Incidents: Realigned vs. Unaltered Exchange Expectations

The possibility of realigning exchange expectations and breaking an existing exchange cycle occurred when entrepreneurs occasionally veered into an alternate interaction ritual or at least had the opportunity to do so. We refer to such moments of possibility as interaction ritual crossover incidents. As we show next, our findings reveal that while intervening bonding rituals may realign exchange expectations set by tournament rituals, expectations set by a bonding ritual chain appear robust to intervening tournament-like rituals.

Discussion of Qualitative Findings

Two key surprising findings emerge from our qualitative analysis. First, by virtue of occupying the open workspace, all entrepreneurs had equal opportunity to interact with each other. Yet interaction patterns, which were comparable in the beginning, diverged considerably across camps. Composed of entrepreneurs who were newcomers to the accelerator, Camps B and C followed similar interaction trajectories initially. But the interaction trajectory of Camp C soon flattened out, whereas Camp B’s interaction patterns quickly came to resemble those of Camp A entrepreneurs who had stayed on from the previous program period (see Figure 2). While a tight-knit network populated by affective ties emerged across Camps A and B, Camp C entrepreneurs not only gradually distanced themselves from their peers in Camps A and B but also discontinued interactions with peers from their own camp. Figures 4a and 4b show cumulative interactions at the end of phase 1 and phase 2, respectively.

OPEN IN VIEWER

Figure 4a.

Cumulative Interactions among Entrepreneurs in Phase 1*

OPEN IN VIEWER

Figure 4b.

Cumulative Interactions among Entrepreneurs in Phase 2*

Second, entrepreneurs across all three camps engaged in unilateral exchanges. Yet while unilateral giving in Camps A and B evolved into a thriving generalized exchange system, unilateral exchanges in Camp C succumbed to the inherent social dilemma resulting in a breakdown of social resource exchange altogether. It is striking that even small acts of giving (e.g., offering a ride home) elicited gratitude among receivers in Camps A and B, while acts of giving that objectively seemed more valuable (e.g., offers to invest) went unacknowledged in Camp C.

Our qualitative analysis allowed us to unpack the mechanisms contributing to the divergence in interaction patterns and exchange outcomes across camps over time. Interaction rituals shaped micro social order indirectly through dyadic exchange processes at the workspace. While expectations of exchange emerged in interaction rituals, it was entrepreneurs’ fulfillment of or failure to realize these expectations that directed their exchange relationships with their peers along divergent paths. Fulfilled exchanges in the workspace set in motion a virtuous cycle of giving and gratitude, marking the emergence of a generalized exchange system. Failed exchanges in the workspace set in motion a vicious cycle of avoidance and shaming, marking the non-emergence of a generalized exchange system.

While our data on tournament rituals were drawn from several sources (i.e., observations, interviews, and social media archives), details on bonding rituals primarily came from interviews and social media archives. Considering that interview data represent retrospective accounts and can be subject to biases, it is possible that entrepreneurs in Camps A and B overstated the importance of bonding rituals in their interviews. Moreover, whereas participation in bonding rituals might be endogenous to interaction dynamics in the workspace, participation in tournament rituals was exogenous as it was mandated by the accelerator. This raises the possibility that the intense workplace interactions observed in Camps A and B might be a result of unobservable factors, which might have contributed to participation in bonding rituals as well. To pursue these issues, we decided to augment our qualitative insights with quantitative analyses. These analyses not only reflect a clear sequence of events over time but also allow us to account for the endogenous nature of bonding rituals and to rule out alternative explanations. In our quantitative analyses, we test (a) whether participation in the two types of interaction rituals systematically influenced workplace interactions and (b) whether workplace interactions indeed mediate the relationship between participation in interaction rituals and the emergence of generalized exchange.

Dependent Variables

Workplace interactions between dyads is a directed and repeatable process. An entrepreneur dyad entered the risk set when the pair was present on a given day and each node belonged to a different start-up. With these restrictions, the 49 entrepreneurs in our sample generated a total of 2,106 directed dyads. In any given day, if founder A initiated at least one conversation with founder B from a different start-up, we coded workplace interaction from founder A to founder B as 1 on that day and as 0 otherwise. Similarly, if founder B initiated at least one conversation with founder A, we coded workplace interaction from founder B to founder A as 1 on that day and 0 otherwise. However, when founders A and B met each other by the watercooler and spontaneously engaged in a conversation, we treated this interaction as bidirectional—i.e., an interaction was recorded for both founders in the dyad. In total, we recorded 938 such daily workplace interactions. While we considered only one directed interaction per dyad and workday in our main analysis, we also ran separate analyses to account for multiple daily interactions per dyad (see the upcoming section “Alternative Explanations and Robustness Checks” for details).

Perception of resource gain between dyads is a directed process. Each dyad of entrepreneurs could have exchanged several resources of varying importance during the study period, but the perception of gain or loss from an exchange depends on the subjective value attached to the social resources given to or received from a peer. Accordingly, each entrepreneur might recall a resource exchange only if he or she has expended considerable effort or incurred opportunity cost in giving it to a peer or has received a social resource with high subjective value from a peer. Thus, similar to Willer, Flynn, and Zak (2012), we collected data on social resource exchanges during exit interviews with each entrepreneur in our study. Specifically, we asked each entrepreneur to list the names of entrepreneurs to whom they gave social resources and the names of entrepreneurs from whom they received social resources, as well as to explain the type of resource in each exchange (for use in our qualitative analysis). We created a measure of perception of resource gain as follows: For any entrepreneur A, we measured their perception of dyadic resource gain vis-à-vis a given peer B as 1 if A recalled receiving social resources from B but did not recall giving any social resource to B. This variable took the value −1 if A did not recall receiving any social resource from B but recalled giving a social resource to B. In all other cases, this variable took the value 0. ⁷

Independent Variables

In our models predicting workplace interaction, we had six key independent variables capturing entrepreneurs’ participation in either tournament or bonding rituals. With regard to participation in tournament rituals, we first included dyad tournament ritual. This variable took the value 1 if both members of a pair attended a social event characterized by tournament rituals (i.e., a mandatory formal weekly meeting in Camp C) in the previous week, and 0 otherwise. Second, we included source only tournament and target only tournament, which took the value 1 if only the source or target node participated in the tournament ritual, and 0 otherwise.

With respect to participation in bonding rituals, we first included dyad bonding ritual, measured as the number of informal social events characterized by bonding rituals that both members of the pair attended together during the weekend (ranging from 0 to 3 in our data). We also included source only bonding and target only bonding, which were measured as the number of bonding rituals in which the source node or the target node alone participated.

In our models predicting perception of resource gain, with regard to tournament rituals, we included three variables similar to the ones mentioned above—cumulative dyad tournament, cumulative source only tournament, and cumulative target only tournament. These variables were measured at the end of the study period and capture the interactions in tournament rituals over the entire study period. With regard to bonding rituals, we included the cumulative number of bonding rituals in which the dyad, the source, and the target participated over the study period (cumulative dyad bonding, cumulative source only bonding, cumulative target only bonding). Finally, cumulative workplace interactions captured the total number of workplace interactions a dyad engaged in over the observation period.

Control variables

We included as controls several network context variables that are widely considered to contribute to tie generation—prior ties, reciprocity, indirect ties, and proximity (Rivera, Soderstrom, and Uzzi, 2010). In addition to these variables, we controlled for several individual- and dyad-level variables to account for alternative explanations, which we discuss in detail in the “Alternative Explanations and Robustness Checks” section. We summarize the measurement of all control variables in Table 3.

OPEN IN VIEWER

Table 3.

Overview of Control Variables

Control Variable	Description
Homophily variables
Status homophily	1 if both members of the pair graduated from similar status universities (e.g., both categorized as either elite or non-elite); 0 otherwise
Start-up experience homophily	1 if both members of the pair have prior experience with founding start-ups; 0 otherwise
Business homophily	1 if both members of the pair operate in the same business space or vertical (e.g., fintech, travel, etc.); 0 otherwise
Network context variables: actor & dyad specific
Out degree source	Total number of outgoing ties for entrepreneur i in time t–1
In degree target	Total number of incoming ties for entrepreneur j in time t–1
Indirect ties	1 if there is at least one indirect tie between entrepreneur i and j in time t–1 through a tie with a common alter
Reciprocity	1 for the pair i→j in time t if there was a directed tie from j→i in time t–1
Prior tie	1 if members of a pair interacted with each other in time t–1
Proximity	Ranges from 1 to 6, with higher scores indicating spatial proximity of the pair. We first coded the adjacency of the table based on the approximate number of tables between founders’ table and the focal table (e.g., tables in opposite corners were assigned a score of 6, as they are farthest away from each other, whereas tables next to each other were assigned a score of 1). To arrive at the proximity scores, we reverse-scored the table adjacency scores so that higher values reflect higher proximity (see Figure 2 for the seating map).
Source characteristics
Source position: CTO	1 if source is a CTO in his/her company; 0 otherwise
Source country: U.S.	1 if source is a U.S. citizen; 0 otherwise
Source identity: Darwinian, missionary	We created three source identity categories—Darwinian, missionary, and communitarian. From these three categories, we generated two dummy variables—Darwinian and missionary—with communitarian being the reference category.
Target characteristics
Target position: CTO	1 if target is a CTO in his/her company; 0 otherwise
Target country: U.S.	1 if target is a U.S. citizen; 0 otherwise
Target identity: Darwinian, missionary	We created three target identity categories—Darwinian, missionary, and communitarian. From these three categories, we generated two dummy variables—Darwinian and missionary—with communitarian being the reference category.

Estimation Techniques

To estimate workplace interactions, we used hazard rate models. These are suitable when analyzing the effect of time-varying covariates on discontinuous events such as tie formation (de Nooy, 2008). We used a Cox conditional gap time (also known as conditional risk set) model for repeated events (Prentice, Williams, and Peterson, 1981) as our primary specification. This model assumes that the duration time resets after each event occurrence. Hence, after the first event occurs, the dyad enters the risk set for the second event. By stratifying on event number and preserving information on the ordering of events, the model allows the baseline hazard function to vary across repeated events while the parameters of covariates stay the same across events (Jones and Branton, 2005). To deal with the non-independence of observations, we used robust standard errors that are clustered on both actors of a dyad simultaneously (Cameron, Gelbach, and Miller, 2011; Dahlander and McFarland, 2013) and employed Kleinbaum, Stuart, and Tushman’s (2013) clus_nway routine in Stata. Please see “Alternative Explanations and Robustness Checks” below for details on additional analyses using alternative specifications.

To estimate perception of resource gain, we used a tobit model. Tobit models are suitable to analyze truncated dependent variables such as ours. As before, standard errors in this model are clustered on both entrepreneurs using the clus_nway routine (Kleinbaum, Stuart, and Tushman, 2013).

Estimates for Workplace Interactions: Effects of Interaction Rituals

Table 4 presents the results of the Cox proportional hazard rate models predicting directed workplace interactions between dyads (i.e., A’s interactions with B and B’s interactions with A). Model 1 is the baseline with all controls. Model 2 shows the effect of participation in tournament rituals on workplace interactions. Models 3 and 4 show the effect of participation in bonding rituals on workplace interactions. Model 5 is the full model with all the controls and the theorized effects included. Below, we only interpret the coefficients of our theorized variables and the coefficients of those control variables that are not further discussed in the alternative explanations section. In line with research on tie formation (e.g., Rivera, Soderstrom, and Uzzi, 2010), proximity (β = .164; SE = .048; p = .001), indirect ties (β = 1.303; SE = .131; p < .001), and prior ties (β = .839; SE = .155; p < .001) explained the likelihood of workplace interaction.

OPEN IN VIEWER

Table 4.

Stratified Cox Gap Time Models for Likelihood of Workplace Interactions*

Covariates	Model 1	Model 2	Model 3	Model 4	Model 5
Network context variables
Proximity	0.164 •••	0.210 •••	0.195 •••	0.198 •••	0.228 •••
	(0.05)	(0.06)	(0.04)	(0.05)	(0.05)
Indirect ties	1.303 •••	1.123 •••	1.292 •••	1.297 •••	1.117 •••
	(0.13)	(0.11)	(0.13)	(0.13)	(0.11)
Reciprocity	0.248	0.244	0.152	0.155	0.158
	(0.26)	(0.25)	(0.23)	(0.23)	(0.21)
Repeated ties	0.839 •••	0.887 •••	0.942 •••	0.947 •••	0.922 •••
	(0.15)	(0.16)	(0.16)	(0.16)	(0.16)
Out degree source	0.087 •••	0.059 •	0.055	0.054	0.041
	(0.03)	(0.03)	(0.03)	(0.03)	(0.03)
In degree target	0.091 •••	0.070 •	0.046	0.048	0.048
	(0.02)	(0.03)	(0.03)	(0.03)	(0.03)
Source characteristics
Source position: CTO	−0.050	−0.107	−0.108	−0.098	−0.139
	(0.22)	(0.17)	(0.20)	(0.20)	(0.15)
Source country: U.S.	−0.773 ••	−0.417 ••	−0.718 ••	−0.684 ••	−0.379 ••
	(0.24)	(0.13)	(0.22)	(0.25)	(0.14)
Source identity: Darwinian	−0.066	0.020	−0.181	−0.171	−0.060
	(0.12)	(0.09)	(0.12)	(0.12)	(0.09)
Source identity: missionary	0.417	0.634 •••	0.282	0.262	0.516 ••
	(0.29)	(0.19)	(0.27)	(0.29)	(0.19)
Target characteristics
Target position: CTO	−0.140	−0.173	−0.170	−0.173	−0.197
	(0.20)	(0.14)	(0.17)	(0.17)	(0.13)
Target country: U.S.	−0.765 ••	−0.469	−0.685 •	−0.687 •	−0.451
	(0.29)	(0.27)	(0.29)	(0.29)	(0.27)
Target identity: Darwinian	−0.023	0.087	−0.121	−0.113	0.007
	(0.11)	(0.10)	(0.11)	(0.11)	(0.10)
Target identity: missionary	0.551 •	0.726 •••	0.379	0.385	0.603 •••
	(0.26)	(0.14)	(0.23)	(0.24)	(0.15)
Homophily variables
Status homophily	0.315 •	0.477 •••	0.290	0.299	0.436 ••
	(0.16)	(0.14)	(0.16)	(0.16)	(0.14)
Startup experience homophily	0.012	0.060	0.044	0.034	0.084
	(0.05)	(0.06)	(0.07)	(0.07)	(0.07)
Business homophily	0.150	0.127	0.094	0.072	0.087
	(0.11)	(0.10)	(0.10)	(0.10)	(0.09)
Bonding ritual residuals
Dyad bonding residual				−0.079	−0.069
				(0.05)	(0.05)
Source only bonding residual				−0.051	−0.043
				(0.06)	(0.06)
Target only bonding residual				0.038	0.028
				(0.02)	(0.02)
Tournament ritual participation
Dyad tournament		−2.369 •••			−2.252 •••
		(0.38)			(0.39)
Source only tournament		−2.084 •••			−2.007 •••
		(0.28)			(0.29)
Target only tournament		−1.441 •••			−1.346 •••
		(0.26)			(0.27)
Bonding ritual participation
Dyad bonding			0.728 •••	0.631 •••	0.449 •••
			(0.08)	(0.12)	(0.09)
Source only bonding			0.236 •••	0.315 •••	0.192 •
			(0.06)	(0.08)	(0.08)
Target only bonding			0.235 ••	0.282 •	0.166
			(0.08)	(0.13)	(0.12)
Number of observations	15,300	15,300	15,300	15,300	15,300
Number of founders	49	49	49	49	49
Number of dyads	2106	2106	2106	2106	2106
Number of interactions	938	938	938	938	938

•

p < .05; ^••p < .01; ^•••p < .001; two-tailed tests.

*

Standard errors are in parentheses (clustered for source, target, and dyad).

Models 2 and 5 show the effect of participation in tournament rituals on workplace interactions. As predicted, the estimated coefficient for the dyad tournament ritual was negative (model 2: β = −2.369; SE = .378; p < .001 | model 5: β = −2.252; SE = .39; p < .001). An examination of this variable’s marginal effect in model 5 demonstrates that there was an 89.5-percent ( e − 2 . 252 − 1 ) decrease in workplace interaction in a dyad in which both the source and target participated in a tournament ritual compared with dyads in which neither participated. Models 3, 4, and 5 show the effect of participation in bonding rituals on workplace interaction. We interpret the results of model 4 here, as it also contains corrections for endogeneity concerns. As expected, the estimated coefficient for the dyad bonding ritual was positive (model 4: β = .631; SE = .116; p < .001 | model 5: β = .449; SE = .092; p < .001). An examination of the marginal effect of this variable in model 5 demonstrates that there was a 56-percent ( e 0 . 449 − 1 ) increase in workplace interaction for each additional bonding ritual in which both the source and target participated in a given period.

Estimates for Perception of Resource Gain: Effects of Rituals and Workplace Interactions

Table 5 presents the results of the Tobit models predicting perception of resource gain of a focal source vis-à-vis a focal target in a focal dyad (i.e., A’s perception of resource gain from B in dyad AB and B’s perception of resource gain from A in dyad BA). Model 1 is the baseline with all controls included. Models 2 and 3 show the effect of the cumulative tournament rituals and cumulative bonding rituals, respectively, on perception of resource gain. Model 4 shows the additive effect of participation in cumulative tournament rituals and cumulative bonding rituals on perception of resource gain. Model 5 analyzes the mediating effect of cumulative workplace interactions. The estimated coefficients of one of the control variables, indirect ties, are negative, indicating that entrepreneurs who continued to have an indirect tie with a peer did not perceive a resource gain. This suggests that in order to receive resources from a peer, an entrepreneur ought to have at least interacted with the peer once.

OPEN IN VIEWER

Table 5.

Tobit Models Predicting Perception of Resource Gain*

	Model 1	Model 2	Model 3	Model 4	Model 5
Network context variables
Proximity	0.006	0.007	0.008	0.007	0.004
	(0.01)	(0.00)	(0.01)	(0.00)	(0.01)
Indirect ties	−0.020	−0.039 ••	−0.015	−0.031 ••	−0.017
	(0.01)	(0.01)	(0.01)	(0.01)	(0.01)
Source characteristics
Source position: CTO	0.019	0.017	0.023	0.020	0.022
	(0.02)	(0.01)	(0.02)	(0.01)	(0.01)
Source country: U.S.	−0.020	0.002	−0.005	0.014	0.007
	(0.02)	(0.02)	(0.02)	(0.02)	(0.02)
Source identity: Darwinian	−0.001	0.010	−0.004	0.010	0.011
	(0.02)	(0.02)	(0.02)	(0.02)	(0.02)
Source identity: missionary	0.010	0.036	−0.004	0.027	0.030
	(0.02)	(0.02)	(0.02)	(0.02)	(0.02)
Target characteristics
Target position: CTO	−0.007	−0.010	0.001	−0.004	−0.002
	(0.02)	(0.01)	(0.01)	(0.01)	(0.01)
Target country: U.S.	−0.035	−0.021	−0.019	−0.012	−0.015
	(0.02)	(0.01)	(0.02)	(0.02)	(0.01)
Target identity: Darwinian	−0.007	0.002	−0.010	−0.003	−0.005
	(0.01)	(0.01)	(0.01)	(0.01)	(0.01)
Target identity: missionary	0.029	0.048	0.012	0.027	0.027
	(0.04)	(0.04)	(0.03)	(0.03)	(0.03)
Homophily variables
Status homophily	0.004	0.006	−0.004	0.001	−0.002
	(0.01)	(0.01)	(0.01)	(0.01)	(0.01)
Startup experience homophily	−0.008	−0.007	−0.007	−0.006	−0.007
	(0.00)	(0.00)	(0.01)	(0.01)	(0.01)
Business homophily	0.010	0.004	0.011	0.008	0.007
	(0.01)	(0.01)	(0.01)	(0.01)	(0.01)
Bonding rituals residuals
Dyad bonding (cumulative) residual			−0.016	−0.013	−0.007
			(0.01)	(0.01)	(0.01)
Source only bonding (cumulative) residual			−0.003	−0.011	−0.010
			(0.01)	(0.01)	(0.01)
Target only bonding (cumulative) residual			0.009	0.005	0.005
			(0.01)	(0.01)	(0.01)
Tournament ritual participation
Dyad tournament (cumulative)		−0.020 •••		−0.018 ••	−0.015 ••
		(0.01)		(0.01)	(0.01)
Source only tournament (cumulative)		−0.012 •••		−0.012 •••	−0.009 •••
		(0.00)		(0.00)	(0.00)
Target only tournament (cumulative)		−0.009 ••		−0.005 •	−0.003
		(0.00)		(0.00)	(0.00)
Bonding ritual participation
Dyad bonding (cumulative)			0.047 •••	0.035 •••	−0.006
			(0.01)	(0.01)	(0.02)
Source only bonding (cumulative)			0.003	0.001	−0.000
			(0.00)	(0.00)	(0.00)
Target only bonding (cumulative)			0.005	0.003	0.002
			(0.00)	(0.00)	(0.00)
Mediator
Workplace interactions (cumulative)					0.026 •
					(0.01)
Constant	0.002	0.049	−0.042	0.018	0.014
	(0.04)	(0.03)	(0.04)	(0.03)	(0.03)
Sigma	0.181 •••	0.175 •••	0.177 •••	0.172 •••	0.171 •••
	(0.03)	(0.03)	(0.03)	(0.03)	(0.02)
Number of observations	2106	2106	2106	2106	2106
Number of founders	49	49	49	49	49

•

p < .05; ^••p < .01; ^•••p < .001; two-tailed tests.

*

Robust standard errors are in parentheses (clustered for source, target, and dyad).

As expected, the estimated coefficient for cumulative tournament rituals was negative (model 2: β = –.020; SE = .006; p = .001 | model 4: β = –.018; SE = .006; p = .001). With every additional tournament ritual in which a dyad participated, the source’s perception of resource gain drops by .020 points. Similarly, as expected, the estimated coefficient for cumulative bonding rituals was positive (model 3: β = .047; SE = .012; p < .001 | model 4: β = .035; SE = .010; p < .001). With every additional bonding ritual a dyad participated in, the source’s perception of resource gain increases by .047 points.

Our qualitative findings indicated that bonding rituals increased workplace interactions, which in turn encouraged generalized exchange among peers. In line with our expectation, the coefficient for cumulative workplace interactions is positive in model 5 (β = .026; SE = .012; p = .033). With every additional workplace interaction by a dyad, the sources’ perception of resource gain increases by .026 points. In this model, the coefficient for cumulative bonding rituals drops out of significance (β = –.006; SE = .017; p = .704), suggesting that workplace interactions fully mediate the relationship between bonding rituals and perception of resource gain. A Sobel test confirmed this mediation effect (t = 2.013; p = .044). There is also some evidence of partial mediation of workplace interactions in the relationship between tournament rituals and perception of resource gain. The coefficient for cumulative tournament rituals remains significant (model 5: β = –.015; SE = .006; p = .008) in the full model, even though it is less pronounced than in model 2 (β = –.018; SE = .006; p = .001). A Sobel test confirmed this mediation effect (t = −2.048; p = .041).

Discussion of Quantitative Findings

Our quantitative findings provide strong support for our qualitative insights. Whereas participation in bonding rituals resulted in a substantial increase in workplace interactions, participation in tournament rituals produced a notable decrease in workplace interactions. Bonding ritual participants were more likely to perceive their exchanges as gainful because bonding rituals encouraged repeated workplace interactions, which is key to social resource exchange in the workplace and the associated perception of resource gain. Conversely, tournament ritual participants were less likely to perceive their exchanges as gainful because tournament rituals discourage repeated workplace interactions.

Our quantitative analysis mitigated some of the weaknesses of the qualitative analysis by showing that even after correcting for endogeneity of bonding rituals and accounting for other factors that might be related to workplace interactions and generalized exchange, our main findings pertaining to interaction rituals are remarkably robust. Overall, our quantitative analysis supports the inductive theorizing that emphasizes how the type of interaction rituals shapes the emergence or non-emergence of generalized exchange through dyadic exchange processes in workplace interactions.

Alternative Explanations and Robustness Checks

We pursued a variety of potential alternative explanations for more stable interaction patterns and a thriving generalized exchange system in Camps A and B compared with Camp C and ultimately concluded that our key findings are robust. First, Camp C might have been predisposed to being more competitive when entering the program. Second, individual founder characteristics or founder identity of Camp C entrepreneurs might have made them less prone to interactions than entrepreneurs in Camps A and B. Third, to address whether other differences across camps explain the contrasting interaction patterns, we compared the spring 2015 cohort (Camps A, B, and C) with similarly structured camps in the fall 2014 cohort (Camps X, Y and AF), for which we had data from preliminary interviews and observations. ⁹

We also ran several sensitivity analyses. First, we tested whether our quantitative results are robust to alternative specifications that consider analysis time to be discrete. Second, we analyzed an event count model to account for all interactions that a dyad engaged in on a given day. Finally, we excluded Camp C from our analysis to ensure that their nonparticipation in weekly bonding rituals was not inflating our results. Tables A1, A2, and A3 in the Online Appendix describe all these additional analyses in detail.

Toward an Interaction Ritual Theory of Generalized Exchange among Peers

While encouraging unilateral giving, accelerators are exemplary mixed-motive contexts that make salient the challenges associated with the emergence and sustenance of generalized exchange among peers. Findings from this study demonstrate how accelerators can serve as platforms for generalized exchange while simultaneously also hindering it. In particular, we draw attention to substantial differences in the success of establishing a community of giving across different camps and cohorts of the same accelerator. The process model that emerged in this study, shown in Figure 5, identifies the exchange expectations generated in interaction rituals as key to shaping the nature of peer relationships and explaining the emergence or non-emergence of generalized exchanges.

OPEN IN VIEWER

Figure 5.

Process Model of Interaction Rituals and Social Resource Exchange

Interaction rituals within emergent social events significantly shaped the purposive act of giving by producing distinct exchange expectations. Whereas shared emotions in bonding rituals evolved into fellow feeling, those in tournament rituals deteriorated into distrust and apathy. In this respect, micro social encounters (Goffman, 1967) were key contributors to the evolution of interaction ritual chains. Expectations of exchange generated in bonding and tournament rituals fed into micro social encounters in the workspace, which in turn shaped the shared emotions in subsequent bonding and tournament rituals. Generalized exchange emerged among peers only when unilateral giving aligned with exchange expectations. Acts of resource seeking contributed as much to the exchange dynamics as acts of resource giving. While unfulfilled resource requests triggered a vicious cycle of avoidance and shaming, fulfilled resource requests initiated a virtuous cycle of giving and gratitude.

Voluntary engagement in bonding rituals was key to the emergence of virtuous exchange dynamics. In the absence of familiarity among entrepreneurs in the initial period, such voluntary participation in bonding rituals was largely shaped by the common focus in the formal events organized by the accelerator early on. While a common focus on building familiarity (e.g., the new entrepreneur orientation event in Camp B) encouraged the emergence of subsequent bonding rituals, a common focus on show of strength (e.g., the new entrepreneur orientation and weekly formal meetings in Camp C) discouraged the emergence of bonding rituals. But our model shows that vicious exchange dynamics set in motion by mandatory tournament rituals can be reversed when consciously coordinated intervening bonding rituals realign exchange expectations (e.g., social events organized by the intern for Camp AF). Our model and theory offer novel insights that challenge and extend existing research on organizational sponsorship, entrepreneur relationships, social resource exchange, and interaction rituals.

Understanding the Black Box of Social Events within Accelerators

Unpacking the interaction dynamics within social events allowed us to identify the conditions under which accelerators evolve into a community of giving (Amezcua et al., 2013: 1636; Grimes, 2018) versus degenerate into an “invitation to exploitation” (Takahashi, 2000: 1107). Recent research on accelerators has documented several potential gains from co-located peers (Cohen, Bingham, and Hallen, 2019), but it is not evident if these gains were a result of willful sharing of resources by peers or an outcome of resource appropriation by receivers. We were able to resolve this ambiguity by closely examining how the nature of relationships among peers evolved over time.

In our study, instrumental relationships among peers evolved into affective relationships or deteriorated into meaningless encounters depending on whether expectations of exchange set in earlier encounters were met or unmet in subsequent encounters. Interaction rituals within social events gave rise to different expectations of exchange depending on entrepreneurs’ common focus and shared emotions in these events. While tournament rituals created strong expectations of direct reciprocity from peers, bonding rituals did not. Accordingly, unilateral giving that emerged among bonding ritual participants did not diverge from their exchange expectations. But owing to reciprocity expectations among tournament ritual participants, giving to peers resulted in perceived costs for resource givers who viewed it as a zero-sum game. Moreover, the source of vicious exchange dynamics among tournament ritual participants lies in forcing entrepreneurs into tournament rituals before they have had the opportunity to develop the familiarity needed to initiate bonding rituals.

Thus our findings suggest that differentiating between social event types within accelerators matters more to our understanding of emergent exchange relationships than previously anticipated in organizational sponsorship research. While organizational sponsorship scholars agree that social events that allow entrepreneurs to network with their peers are a key feature of start-up accelerators (Amezcua et al., 2013; Cohen, Bingham, and Hallen, 2019), what transpires in different types of social events has largely remained a black box. Our study underscores the importance of accounting for the interaction rituals in different types of social events in accelerators, without which our understanding of exchange dynamics in peer relationships will be incomplete.

Understanding Affective Giving among Strategic Resource Seekers

Our study extends recent efforts to integrate affective processes into entrepreneurship research, especially as they pertain to entrepreneur relationships (e.g., Huang and Knight, 2017). With much emphasis on entrepreneurs’ instrumental resource-seeking behavior in prior entrepreneurship research, entrepreneurs’ affective giving behavior, which lays the foundation for the emergence of cooperative communities, has received only scant attention (see Saxenian, 1994, for an exception). Given that acquiring external resources is a fundamental challenge facing early-stage entrepreneurs, scholars have mainly portrayed entrepreneurs as strategic actors who purposefully use narratives and symbolic means to acquire resources for their ventures (Martens, Jennings, and Jennings, 2007). Peer relationships are no exception to this strategic resource-seeking behavior. For instance, in their work examining relationships between new ventures in the solar energy sector, Hannah and Eisenhardt (2018: 3182) showed that new ventures either competed heavily with other ventures or balanced competition and cooperation in a way that “impose[d] extractive terms” on their partners.

Based on these insights from prior research, what may be more surprising than the eventual collapse of exchange in Camp C is the emergence of generalized exchange in Camps A and B. Considering that entrepreneurs are often conceptualized as instrumental resource seekers, Camps A and B being populated with entrepreneurs who engaged in frequent affective resource giving was contrary to expectations. Entrepreneurs across all camps sought resources, but willing givers emerged only when actors were made to realize their common fate as entrepreneurs thanks to their participation in bonding rituals. While our findings are not inconsistent with the notion that entrepreneurs are instrumental resource seekers (Martens, Jennings, and Jennings, 2007; Stuart and Sorenson, 2007), they bring forth the novel insight that participation in rituals that reinforce expectations of unilateral exchange can nonetheless give rise to affective giving. Further, resource seekers benefitting from their giving peers can trigger a giving–gratitude cycle resulting in more unilateral giving to other peers.

Through this novel peer-to-peer community perspective, our study complements earlier research on the affective relationships between entrepreneurs and potential investors (Huang and Knight, 2017). While this earlier research shed light on how building an affective relationship with a targeted other enables bilateral resource exchange, our findings suggest that for the emergence of a generalized exchange system, affective attachment to the group may matter more than affective attachment to any particular actor. In our study, affective attachment to a group was made possible by expectations of unilateral exchange generated in bonding rituals, followed by a cycle of giving and gratitude triggered by fulfilled exchanges. In contexts such as accelerators, which encourage generalized exchange among peers, resource-seeking attempts become fulfilled exchanges only when resource seekers find affective givers who identify with the group and give without expectation of direct reciprocity.

Understanding Stability and Instability of Generalized Exchanges

Our study provides novel insights into the stability or instability of generalized exchange systems that are at the heart of social exchange theory. We advance an argument that emphasizes the need for an alignment between actors’ exchange expectations and the emergent exchange structure. Only when such alignment is achieved will actors willingly participate in that form of exchange. But exchanges fail when the emergent exchange structure does not align with exchange expectations. The notion that participation in a certain type of exchange is driven by differences in expectations of reciprocity has been missing in prior research, possibly because social exchange theory has predominantly been studied using experiments that assign actors to exchange structures (Molm, Collett, and Schaefer, 2007; Lawler, Thye, and Yoon, 2008, 2009), as though exchanges occur independent of the situations in which they are embedded. Contrary to this assumption, we find that social exchanges are embedded in a chain of social situations in which expectations gradually emerge and become reified. Our findings suggest that understanding how social situations actors find themselves in give rise to different exchange expectations is key to explaining why some generalized exchange systems thrive while others whither or fail to take off.

Delving deeper into the dynamics surrounding failed vs. successful exchanges, our findings support the notion that an affective orientation and pro-group sentiments can be critical (Molm, Collett, and Schaefer, 2007; Willer, Flynn, and Zak, 2012)—but these processes need not be inherent in the exchange structure, as previously assumed. Claims in prior social exchange research that the act of unilateral giving communicates affective orientation to the recipient were not uniformly supported in our study. In fact, acts of unilateral giving can also be perceived as incomplete exchanges awaiting direct reciprocity. Although unilateral giving occurred in all three camps, it communicated affective orientation only in Camps A and B while carrying expectations of delayed reciprocity in Camp C. This finding provides insight into variation within (rather than across) exchange structures, thus helping to explain divergent findings in prior research (e.g., Molm, Collett, and Schaefer, 2007; Lawler, Thye, and Yoon, 2008). Specifically, not all generalized exchange structures are ripe with affective orientation and pro-group sentiments; rather, these processes emerge only when unilateral giving is perceived as such and not as a dyadic reciprocal exchange attempt.

In addition to resource giving, our findings also emphasize the critical importance of resource requests for the emergence of successful generalized exchange. In particular, unilateral exchanges were perceived as beneficial only when the giving was commensurate with what was requested. And unmet resource requests in failed exchanges led rejected actors to attribute the failure to the group and disengage from subsequent interactions and acts of giving. This process underscores the importance of taking into account the outcomes of resource requests, which are yet to receive attention in social exchange research.

Overall, our analyses extend social exchange theory’s insights on the functioning of different types of exchange structures by highlighting their interactional components. Our model emphasizes that not only are micro interactions shaped by exchange structures, but exchange structures are also created, sustained, or altered by micro interactions (Turner, 2019; Zucker and Schilke, 2020), thus adding much-needed process insights into how different types of exchange unfold over time.

Understanding Interaction Rituals in Mixed-Motive Contexts

Research on interaction rituals tends to emphasize how a common focus and intensely shared emotions, which are the key ingredients in rituals, will unify participants (Durkheim, 1912; Collins, 2004; Metiu and Rothbard, 2013). By studying the emergence of interaction ritual chains in a mixed-motive context, we show that the same ingredients that can unify participants—a common focus and intensely shared emotions—may also disrupt social order, depending on their type. Most prior interaction ritual research has treated rituals as a unitary concept. Our study adds nuance by emphasizing that the nature of rituals can differ drastically across bonding and tournament rituals and that such differences have fundamental downstream consequences.

One might have expected that the intense emotions of tension and anticipation that participants in our study shared in early tournament rituals should have brought about an awareness of the common challenges they face as entrepreneurs. But because these shared emotions were founded on a common focus of show of strength, participants primarily paid attention to the costs of cooperation, which generated a strong self-orientation following the interaction ritual, despite intensely shared emotions. Micro social encounters in the workspace following the early tournament rituals became social situations in which entrepreneurs strived to benefit from their peers. This fact suggests that in mixed-motive contexts, a common focus and shared emotions can also make the competitive elements in people’s relationships more salient.

Our findings also shed new light on how interaction rituals emerge and evolve over time resulting in a stable or unstable social order. Interaction ritual theory argues that collective sentiments toward the group will fade away unless they are renewed and shared emotions are recreated in subsequent group rituals (Collins, 2004). Our findings show that shared emotions are not only renewed, but they can also either intensify or deteriorate as interaction rituals evolve over time, giving rise to a stable or unstable social order. In this respect, micro social encounters à la Goffman (1967) are key contributors to the evolution of interaction ritual chains. Expectations of interaction generated in bonding and tournament rituals feed into micro social encounters at the workspace, which in turn shape the shared emotions in subsequent bonding and tournament rituals. When expectations are met in dyadic micro social encounters, positive shared emotions intensify in subsequent meso group rituals, giving rise to a stable social order. Conversely, when expectations remain unmet in dyadic micro social encounters, shared emotions in meso group rituals can deteriorate, resulting in an unstable social order.

Our study also emphasizes that individuals can be engaged in multiplex interaction rituals with each other, which can make the interaction dynamics among participants more complex. In mixed-motive contexts, the common focus in one interaction ritual might contradict the common focus in another. In our study, Camp AF in fall 2014 began with tournament rituals, as did Camp C in spring 2015. But in Camp AF, intermittent bonding rituals organized by an intern prevented the shared emotions in the tournament rituals from spiraling downward. The positive emotions in bonding rituals introduced by the intern could override the negative emotions in tournament rituals, realigning expectations and reviving a stable social order.

Future Research and Generalizability to Other Settings

Our findings highlight how interaction rituals can shape the (non-)emergence of generalized exchange among entrepreneurs. But owing to the short period of time entrepreneurs spend in an accelerator, it would be premature to conclude that the generalized exchange that emerges in accelerators represents enduring communities of giving. Initial evidence from our study suggests that accelerators sow the seeds of generalized exchange by enabling a chain of bonding rituals among peers, which entrepreneurs may take with them to other contexts. The first author later interviewed the founder of one of the start-ups that graduated from the accelerator we studied. This start-up grew considerably in the five years after it graduated from the accelerator; at the time of the follow-up interview, it was big enough to implement generalized exchange norms inside the organization of 65 employees. In this company, Ethan and his co-founders assign economic goals such as “client acquisition targets” that can induce competition to the higher levels of management. At the employee level, they discourage people from “elbowing their way forward” and set examples of cooperation by engaging in bonding rituals with employees. “As it is, clients give us a tough time, and working with them is intense enough. At least we should have fun when we work with each other,” Ethan said. “I frequently have beer or lunches with our employees to bond with them. The general vibe around here is not about chasing individual goals but working together to reach our company’s goals.” Future research should examine further whether participants of bonding rituals carry the seeds of generalized exchange sown in platforms such as accelerators to other contexts, such as the design of their own organizations. Similarly, recent research has shown that organizations’ moral behavior is shaped by routine interactions with socially responsible peers (Krishnan and Kozhikode, 2015). Future research might examine whether the source of such positive influence lies in the bonding rituals in these interactions.

Our findings have broader implications for a variety of social settings beyond accelerators, especially those in which actors forming a cohort are expected to both cooperate and compete with each other—for example, faculty competing for grants and employees within teams competing for promotions (Johnson et al., 2006; Srivastava et al., 2017). We see the accelerator setting as a conservative context in which to investigate the cooperative–competitive processes for two reasons. First, unlike in firms, where teams are forced to cooperate through fiat in spite of competition to achieve tasks, entrepreneurs in accelerators can only be encouraged to cooperate. Second, entrepreneurs are in accelerators for a limited period and are under pressure to make the most of this time. Unlike teams, whose success hinges on cooperation, success for entrepreneurs participating in an accelerator program critically depends on the progress of their individual start-ups. Hence cooperation among entrepreneurs cannot be taken for granted, making peer relationships within accelerators a conservative test of the positive dynamics spawned by bonding rituals. Future research might examine the role of interaction rituals in facilitating peer acceptance in localized networks, which is shown to be vital for cultural assimilation in organizations (Srivastava et al., 2017). Participation in distinct interaction ritual types—tournament vs. bonding—during team formation might shed light on how peer acceptance originates in localized networks in organizations.