54
Conversational Peers and Idea Generation: Evidence from a Field Experiment * Sharique Hasan Stanford GSB Rembrand Koning Stanford GSB September 29, 2015 Abstract Social interaction is thought to affect individual and organizational innovation. We argue that individuals and teams are better able to generate high quality ideas if they converse with peers who provide them with new insight, perspective and information. Further, we argue that not all individuals can equally capitalize on this new information. Specifically, extroverted peers, because they are more willing to share and transmit their experiences facilitate idea generation. Moreover, innovators who are open to experience absorb this new information better and can more effectively convert it to better ideas. We test our claims using novel data from a randomized field experiment at an entrepreneurship bootcamp where individuals were randomized to both teams and conversational peers. We find that conversations with extroverted peers help individuals generate higher quality ideas, more open individuals benefit more from such peers, and teams with more cohesion can convert these raw ideas into better performance. * Both authors contributed equally. Manuscripts from this project have alternating authorship sequence. Please direct correspondence to [email protected] and [email protected]. Special thanks to everyone who made Innovate Delhi happen, especially Ponnurangam Kumaraguru and Randy Lubin. This research was made possible by IIIT-Delhi and the Indian Software Product Roundtable. This work was funded by Stanford’s Graduate School of Business, Institute for Research in the Social Sciences, and the Institute for Innovation in Developing Economies. 1

Sharique Hasan Rembrand Koning Stanford GSB …...Sharique Hasan Stanford GSB Rembrand Koning Stanford GSB September 29, 2015 Abstract Social interaction is thought to a ect individual

  • Upload
    others

  • View
    7

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Sharique Hasan Rembrand Koning Stanford GSB …...Sharique Hasan Stanford GSB Rembrand Koning Stanford GSB September 29, 2015 Abstract Social interaction is thought to a ect individual

Conversational Peers and Idea Generation:Evidence from a Field Experiment ∗

Sharique HasanStanford GSB

Rembrand KoningStanford GSB

September 29, 2015

Abstract

Social interaction is thought to affect individual and organizational innovation. We arguethat individuals and teams are better able to generate high quality ideas if they converse withpeers who provide them with new insight, perspective and information. Further, we arguethat not all individuals can equally capitalize on this new information. Specifically, extrovertedpeers, because they are more willing to share and transmit their experiences facilitate ideageneration. Moreover, innovators who are open to experience absorb this new informationbetter and can more effectively convert it to better ideas. We test our claims using novel datafrom a randomized field experiment at an entrepreneurship bootcamp where individuals wererandomized to both teams and conversational peers. We find that conversations with extrovertedpeers help individuals generate higher quality ideas, more open individuals benefit more fromsuch peers, and teams with more cohesion can convert these raw ideas into better performance.

∗Both authors contributed equally. Manuscripts from this project have alternating authorship sequence. Pleasedirect correspondence to [email protected] and [email protected]. Special thanks to everyone who madeInnovate Delhi happen, especially Ponnurangam Kumaraguru and Randy Lubin. This research was made possibleby IIIT-Delhi and the Indian Software Product Roundtable. This work was funded by Stanford’s Graduate Schoolof Business, Institute for Research in the Social Sciences, and the Institute for Innovation in Developing Economies.

1

Page 2: Sharique Hasan Rembrand Koning Stanford GSB …...Sharique Hasan Stanford GSB Rembrand Koning Stanford GSB September 29, 2015 Abstract Social interaction is thought to a ect individual

Introduction

What role does social interaction play in the idea generation process? A growing literature in

sociology and management has begun to explore this question, with the aim of learning how

networks facilitate idea generation (Fleming, Mingo and Chen, 2007; Burt, 2004, 1992; Reagans

and Zuckerman, 2001; Reagans, Zuckerman and McEvily, 2004; Tortoriello and Krackhardt,

2010; Azoulay, Graff Zivin and Wang, 2010). One major strand of this literature has focused on

the importance of collaboration. Research unequivocally suggest that teams and co-authorships

dominate lone individuals in the development of new ideas by facilitating recombination of

diverse and specialized knowledge (Singh and Fleming, 2010; Wuchty, Jones and Uzzi, 2007;

Girotra, Terwiesch and Ulrich, 2010). However, co-creation and formal collaboration constitute

only one class of social interaction. Innovators also interact with, and benefit from, more in-

formal interaction with people outside their team (Lee, Lee and Pennings, 2001; Menon and

Pfeffer, 2003). Such informal interactions include conversations with peers in other firms, cus-

tomers, and other individuals not formally involved in their innovation process. Indeed, much of

modern entrepreneurship education and practice encourages conversations to enhance individual

creativity (Blank, 2013; Kelley and Kelley, 2013; Ries, 2011; Sutton and Hargadon, 1996) . As

a result, an increasingly important question is determining whether and why social interaction

affects ideation.

An important source of external knowledge for both innovators and teams are the conver-

sations that they have with external peers (Shue, 2013; Uzzi, 1996, 1997). Peers possess many

important resources and traits that should benefit the innovator. Peers possess general and

specialized knowledge (Sacerdote, 2001; Hasan and Bagde, 2013), diverse contacts (Hasan and

Bagde, 2015), and financial resources (Shane and Cable, 2002). Peers also possess intangible

information accumulated from their own experience that may give the individual new perspec-

tive and insight (Lerner and Malmendier, 2013; Nanda and Sørensen, 2010). By conversing

with an external peer, the focal innovator can develop awareness about new problems, insight

into possible solutions, or develop novel hypotheses by recombining divergent experiences. The

ability of the innovator to generate new insight and perspective from her conversations, and thus

generate good ideas, should rest on the interaction dynamics with her peers (Carrell, Sacerdote

and West, 2013; Hasan and Bagde, 2013). Thus, a key question is understanding the social

factors that lead to more fruitful interaction with conversational peers.

2

Page 3: Sharique Hasan Rembrand Koning Stanford GSB …...Sharique Hasan Stanford GSB Rembrand Koning Stanford GSB September 29, 2015 Abstract Social interaction is thought to a ect individual

An important, though often overlooked, trait likely to enhance interaction dynamics between

the innovator and her peer is the peer’s willingness to share her experiences (Goh, 2002). A large

body of literature in psychology and management suggests that this willingness to talk, share,

and express one’s views can be succinctly captured by a person’s level of extraversion (De Vries,

Van den Hooff and de Ridder, 2006; Cheng-Hua et al., 2007; Matzler et al., 2008). Extraversion is

a relatively stable and independent personality trait (Rammstedt and John, 2007). Extroverts

enjoy and gain energy from social interactions. In conversations, they are talkative, express

positive emotion, and willingly disclosing information about themselves (John and Srivastava,

1999; McCrae and John, 1998). In contrast, introverts are more reserved and reflective, gaining

energy through time alone (Brebner and Cooper, 1978). Conversations, particularly short ones,

with extroverted peers should produce interactions with more informational volume, greater

affect, and are also more likely to expose the listener to the perspectives and experiences of her

peers (McCroskey and Richmond, 1995). Thus, relative to having a more introverted peer, an

innovator should accumulate a greater range of raw material for her idea generation process,

and thus develop better ideas, when she talks to an extrovert (Aral and Van Alstyne, 2011).

Recent research in sociology and economics, however, suggests that the mere presence of a

peer with beneficial traits does not invariably lead to good outcomes for the focal individual

(Carrell, Sacerdote and West, 2013; Hasan and Bagde, 2013). Specifically, research suggests

that a complementary “match” between the traits of the focal individual and peer moderate

the degree of social influence. Absent a match, resources, knowledge, and insight may be lost

in transmission (Beshears et al., 2015; Hasan and Bagde, 2013). In the context of peers, the

focal individual must both absorb the information produced by her peer and and make novel

associations between this information and her own experiences to develop good ideas (Zahra and

George, 2002; Cohen and Levinthal, 1990; Borgatti and Cross, 2003; Liao, Fei and Chen, 2007).

One individual level trait that captures this receptivity to external stimuli—such as information

gleaned from conversations with others—is a person’s openness to experience (hereon, openness)

(John and Srivastava, 1999). Individuals with greater openness possess intellectual curiosity,

openness to sensory experiences, active imaginations, a greater need for cognition, and are

more responsive to engaging stimuli (George and Zhou, 2001; McCrae and Sutin, 2009; McCrae,

1987). Consequently, if conversations with external peers are conceptualized as an information

transmission process, more open innovators should have greater sensitivity to information from

more extroverted peers. Thus, the pairing of an innovator open to experience and an extroverted

3

Page 4: Sharique Hasan Rembrand Koning Stanford GSB …...Sharique Hasan Stanford GSB Rembrand Koning Stanford GSB September 29, 2015 Abstract Social interaction is thought to a ect individual

peer should help the former develop better ideas.

The processes described above outlines the idea generation process of individuals. However,

this process should also hold at the level of the team (Reagans and Zuckerman, 2001; Rea-

gans, Zuckerman and McEvily, 2004; Oh, Chung and Labianca, 2004; Oh, Labianca and Chung,

2006). If members of a team talk to many extroverts, then that team should also possess a

greater quantity of raw material for the idea generation process and also higher quality raw

ideas generated by its individual members. Furthermore, if the team’s members are open to

experience, the quality of their ideas in aggregate should also be superior to teams with less

open members. This simple process of aggregation should result in higher quality ideas for

teams with more open team members and more extroverted peers. Moreover, internal team

dynamics—particularly the extent to which the team members exhibit cohesive and collabora-

tive behaviors—should moderate whether the team can capitalize on the knowledge generated

through external interaction (Reagans, Zuckerman and McEvily, 2004; Woolley et al., 2010).

Despite the theorized importance of external peer conversations for the innovation process,

testing whether peers help individuals and teams innovate is difficult for several reasons. Pri-

mary among these reasons is that while team-memberships, co-authorships, and other formal

collaborations are easier to observe by the researcher, short external conversations are often

unobserved. Thus, external peer conversations are generally outside the purview of empirical

analysis. Second, even if conversations were to be enumerated, capturing the psychological and

other characteristics of peers (e.g. extraversion or openness) is necessary to test our proposed

extraversion-openness match theory. Third, to test whether individuals and teams have better

ideas because of these conversations, researchers should have access to the wide set of ideas

generated at both the earlier and later stages of the process as well as ratings of such ideas in

terms of their quality. Fourth, even if the above data were available, the selection of conver-

sational peers by the focal innovator in most contexts is endogenous—depending on the often

unobserved traits and preferences of both the focal individual and the peer (Sacerdote, 2001;

Manski, 1993, 2000). Thus, without accounting for self-selection in peer choice, the relationship

between peers and ideation outcomes may be spurious. Thus, the ideal test would be one where

there is random assignment of conversation partners. Finally, to assess the impact of conversa-

tions at the team level, the composition of the team based on the openness of its member should

not depend on the self-selection of teams premised on the value of openness (Hartmann et al.,

2008). Thus, teams ideally should also be randomly composed such that variance in a team’s

4

Page 5: Sharique Hasan Rembrand Koning Stanford GSB …...Sharique Hasan Stanford GSB Rembrand Koning Stanford GSB September 29, 2015 Abstract Social interaction is thought to a ect individual

openness and members’ cohesive orientations is not due to self-selection.

To overcome these empirical challenges we designed a novel field experiment embedded in

a bootcamp for young entrepreneurs held in New Delhi, India in June of 2014. During the

first week of the bootcamp the 112 participants were randomly assigned to work in 40 teams

of approximately three individuals to develop a mobile software product concept for the Indian

wedding industry. To overcome the four empirical hurdles described above we measured every

individual’s and team’s idea generation and development process during the week, ranging from

the generation of the original concept by individual team members to the team’s submission

of the final product concept and prototype for prizes at the week’s end. Furthermore, we

measured multidimensional evaluations of the ideas generated at each stage. To ensure that we

could get valid causal estimates of peers on idea quality, we randomly assigned individuals to

three external conversation peers, with each conversation lasting 20 minutes, on the second day

of this competition. We used pre-event measures of peers’ and individuals’ extraversion and

openness to test our hypotheses about the role of peers in the idea generation process.

Our results provide strong support for the hypothesis that conversations with peers improve

a focal individual and team’s ability to generate novel and high quality ideas. Specifically our

results indicate that individuals who have conversations with extroverted peers have higher

quality ideas in terms of business value, demand potential and novelty. Furthermore, we find

that the quality of an individual’s ideas improve even more when a match exists between a

high-openness focal innovator and an extroverted peer. Finally, we find that teams who, in

aggregate, had more extroverted conversation peers had better rated final projects and that this

effect is greater for teams with members high in openness to experience and who show greater

team cohesion.

The theory and empirical results provided in this article extend the sociological model of

innovation in several ways (Reagans and Zuckerman, 2001). We show that external peers influ-

ence the idea generation process for individuals and also aggregate up to the level of the team.

Second, our results dovetail with a growing body of literature that suggests that peer-influence is

not a homogenous process, but fundamentally depends on a match between peer traits and focal

and team level characteristics, in our case peer extraversion and individual and team openness

(Carrell, Sacerdote and West, 2013). Third, our model tests a key idea in the literature that

internal team processes affect the value of external knowledge. This component of our model

allows us to enrich the standard sociological account by embedding within it psychological and

5

Page 6: Sharique Hasan Rembrand Koning Stanford GSB …...Sharique Hasan Stanford GSB Rembrand Koning Stanford GSB September 29, 2015 Abstract Social interaction is thought to a ect individual

cognitive processes. We think many of the insights from our study should be useful for scholars

who study social processes beyond innovation, such as educational success, labor market search,

and the development of cultural tastes. Finally, a key contribution of this research is method-

ological. We provide a novel field-experimental framework for studying the dynamics of social

influence and innovation in a rigorous manner.

Theory and Hypotheses: Social Dimensions of Ideation

Social factors are central to how sociologists as well as practitioners conceptualize the produc-

tion function of innovation (Fleming, Mingo and Chen, 2007; Burt, 2004, 1992; Reagans and

Zuckerman, 2001; Reagans, Zuckerman and McEvily, 2004; Tortoriello and Krackhardt, 2010;

Azoulay, Graff Zivin and Wang, 2010). In the development of new ideas, products, and ventures,

research has found that individuals and teams with a wide network of external contacts are ad-

vantaged over their equally capable, though less connected, competition (Fleming, Mingo and

Chen, 2007; Burt, 2004). Networks are thought to provide innovators with a host of benefits,

ranging from greater access to capital (Shane and Cable, 2002), an edge in hiring talent (Fer-

nandez and Fernandez-Mateo, 2006), as well as status and reputation (Podolny, 2001, 1993).

Perhaps the most intriguing of the network-based mechanisms is the idea that external networks

lead to a more regular and higher fidelity stream of raw information (Burt, 2004; Reagans and

Zuckerman, 2001). Networks are thought to aid learning on a range of dimensions. Better

networked innovators can acquire more information about the problems of customers, new tech-

nologies, and accumulate diverse facts and insights that they can combine into novel hypotheses.

All these benefits give the networked individual an advantage in the innovation process (Burt,

2004).

Yet, this network effect rests on several critical assumptions about the concrete social in-

teractions that individuals have with members of their network (Reinholt, Pedersen and Foss,

2011). First, the network effect assumes that the focal individual actually activates his or her

network (Smith, Menon and Thompson, 2012; Smith, 2005). For the network to be a source of

information, the focal innovator must engage and converse with her contacts or peers,. Through

informal chats, lunches, feedback sessions and the like she can learn about the experiences,

insights, and knowledge of her peers. Second, even if networks are activated and interaction

occurs, the ability of the focal individual to learn something through the interaction depends

6

Page 7: Sharique Hasan Rembrand Koning Stanford GSB …...Sharique Hasan Stanford GSB Rembrand Koning Stanford GSB September 29, 2015 Abstract Social interaction is thought to a ect individual

fundamentally on whether her peer possesses resources such as higher stocks of beneficial expe-

riences, insight and knowledge (Sacerdote, 2001; Zimmerman, 2003). Resource-rich peers should

have a far greater ability to transmit important information to the innovator than resource-poor

ones. However, research in a variety of contexts suggests that even when concrete interaction

between an individual and her peer occurs and a peer possesses important resource stocks, the

magnitude of peer effects are often small and insignificant (Sacerdote, 2014). A fundamental

question arises: when does interaction lead to transmission?

In recent years, scholars have suggested that one approach to understanding when peers

will matter is to view interaction as a noisy medium for resource and information transfer

(Carrell, Sacerdote and West, 2013). Even if two individuals interact and the peer possesses

valuable information, characteristics of the dyad or interaction may either facilitate or hinder

transfer and thus affect the whether the focal individual benefits from the interaction (Diette

and Oyelere, 2014). On the one hand, the transfer of information from a peer depends on

that peer’s willingness and ability to share her experience, perspective, or knowledge (De Vries,

Van den Hooff and de Ridder, 2006). If the peer does not share (either by choice or by habit)

the resources she possesses, then the focal individual will not benefit. On the other hand, even if

the peer does share information, a beneficial transfer may not occur because the focal individual

is either unwilling or incapable of absorbing the shared knowledge (Tortoriello, 2015). Thus,

the extent to which peer interaction is beneficial to an individual depends on a peer’s ability to

share knowledge, the focal individuals ability to absorb it, and therefore a match between the

sender and receiver’s traits.

Peer Extraversion and Idea Generation

While peer interaction may provide a wide range of benefits to individuals, perhaps the most im-

portant interaction in the early days of the ideation process is informal conversation (Sosa, 2011;

Perry-Smith, 2006). Conversations—over lunch, coffee, or in other contexts—is commonplace.

Individuals, during the idea generation stage, meet with friends, co-workers, acquaintances and

even the acquaintances of their acquaintances to seek advice, share a raw idea, or just chat.

For example, an entrepreneur who is developing a human resource management product may

meet with a CEO of a rapidly growing firm, talk to a friend about her experience as an em-

ployee, or seek advice from an acquaintance who is a human resources professional. Through

7

Page 8: Sharique Hasan Rembrand Koning Stanford GSB …...Sharique Hasan Stanford GSB Rembrand Koning Stanford GSB September 29, 2015 Abstract Social interaction is thought to a ect individual

these interactions, the entrepreneur gains insight into important though overlooked problems,

potential solutions, and ultimately business opportunities. Indeed, given the potential for early

conversations to yield novel ideas, many modern approaches to innovation such as the “design

thinking” methodology of IDEO and the “lean startup” methodology rely on early conversa-

tions to guide product strategy (Kelley and Kelley, 2013; Ries, 2011; Blank, 2013; Brown et al.,

2008). The purpose of these conversations is not to acquire specialized knowledge about a tech-

nology. Rather, these early conversations help the entrepreneur to accumulate a wide base of

raw data that can be used to generate ideas and develop high potential business plans. A more

voluminous amount of data should result in greater pool of raw material.

What then should be the traits of an ideal conversation at this stage? More ideal conversation

should yield specific and authentic descriptions of experiences, detailed anecdotes, emotions, as

well as the peers’ opinions about problems and even solutions (Kelley and Kelley, 2013; Ries,

2011). Conversely, less ideal conversation at this stage would be slow, lack description of concrete

experiences, and would possess little of the emotional content that could enhance the likelihood

that the innovator experiences her conversational partners perspective (Kelley and Kelley, 2013).

Thus, after the latter conversation the innovator may leave with a few facts about her peer, but

no deep insight into someone else’s experiences and emotions with a given problem domain. The

former conversation, however, should result in the innovator having more raw information for

idea generation.

Consider an entrepreneur designing a product for parents traveling with young children. If

the entrepreneur is fortunate enough to talk to a parent willing to share his experiences in detail,

she could learn a range of information: the difficulty in installing and removing a car seat from

a car, the difficulty in finding a place to change a diaper, the difficulty of taking a stroller up a

tall flight of stairs, and the need to hold a baby close when she is crying. On the other hand, if

an entrepreneur talks to a parent less forthcoming with his experience, she may only learn that

traveling with children is hard, but few specific points of frustration or work-arounds developed

by parents. Thus, the former entrepreneur—the one who converses with a parent willing to

share the nuances of his experience—is better positioned to develop a higher quality and more

useful product. This is because she has access to detailed information about the process she is

trying to innovate upon.

One dimension of a conversational peer that can produce variance in the amount and rich-

ness of information is a peer’s level of extraversion (De Vries, Van den Hooff and de Ridder,

8

Page 9: Sharique Hasan Rembrand Koning Stanford GSB …...Sharique Hasan Stanford GSB Rembrand Koning Stanford GSB September 29, 2015 Abstract Social interaction is thought to a ect individual

2006; Cheng-Hua et al., 2007; Matzler et al., 2008). Extraversion is a stable and independent

personality trait and one of five traits that psychologists believe constitute important variance

in human personality (Rammstedt and John, 2007). Extroverted individuals exhibit a range of

outwardly focused behaviors. In conversation, they are warm, gregarious, talkative, expressive

and show comfort in large groups and around new people (John and Srivastava, 1999; McCrae

and John, 1998). At the opposite pole of this spectrum are introverts. Research finds that

introverted individuals are more likely to keep their emotions private and display quiet and

reserved behavior, especially in the company of new people (Brebner and Cooper, 1978). Inter-

actions and conversations with extroverts are likely to information rich and emotional powerful,

whereas conversations with introverts are likely to possess fewer of these traits. Conversations

with extroverts should produce a greater amount of raw data such as experiences, emotions, and

anecdotes drawn from personal experience of the peer (De Vries, Van den Hooff and de Ridder,

2006). Conversations with introverts are likely to lack the raw volume of experience and emotion

useful at this stage of the ideation process. Thus, we hypothesize:

Hypothesis 1 Individuals who converse with extroverted peers outside their team will develop

higher quality ideas.

Openness, Receptivity and Asymmetry in Ideation

Incorporating a peer’s ability and willingness to share experiences resolves one important bot-

tleneck in the transmission process. However, as with any transmission process the absorptive

ability of the receiver will affect whether transfer is successful (Cohen and Levinthal, 1990; Liao,

Fei and Chen, 2007; Tortoriello, 2015; Tsai, 2001; Zahra and George, 2002). Thus, the ability of

a receiver to absorb and then convert the information she receives from her peer into high quality

ideas may constitute another bottleneck in the innovation process. Two facets of an innovators

behavior are likely to modulate whether she can benefit from the information she receives from

her peer. The first set of behaviors that should affect whether an individual benefits from such

an interaction is whether the innovator is curious and attentive. The focal individual enacts

this behavior through a greater willingness to see others’ perspectives as well as an inquisitive

orientation (Kelley and Kelley, 2013; Csikszentmihalyi, 1997; Grant and Berry, 2011). A second

trait likely to increase the value of external information is a focal individuals ability to make

novel associations using the received stimuli. Combined, these two traits should increase the

9

Page 10: Sharique Hasan Rembrand Koning Stanford GSB …...Sharique Hasan Stanford GSB Rembrand Koning Stanford GSB September 29, 2015 Abstract Social interaction is thought to a ect individual

likelihood that raw information gleaned from conversation with extroverted peers translates into

better ideas.

An individual characteristic that captures such an ability is an innovator’s openness to

experience (George and Zhou, 2001; McCrae, 1987; McCrae and Sutin, 2009). Openness, like

extraversion is one of the core personality traits in the five-factor model (John and Srivastava,

1999). Individuals who have greater levels of openness exhibit many characteristics that allow

them to both better appreciate and absorb external stimuli—such as conversations—as well as

combine the information accumulated through this stimuli into novel ideas (George and Zhou,

2001). In one respect, more open individuals are more intellectually curious and willing to see

things from others’ perspectives. These traits are likely to pair well with extroverted peers

willing to share experiences and perspectives. Conversely, individuals more closed to experience

exhibit fewer of these traits and thus may not be able to capitalize as effectively in conversations

with peers who provide novel insight.

Given the potential variability in both a peer’s ability to transmit information and an in-

novators ability to absorb and use it, it is likely that only a subset of peer interactions are

beneficial in the innovation process. Thus, for effective transmission, a complementary match is

necessary between the sender and receiver. Given a peer willing to talk and share information,

that information is useful for the innovation process insofar as the innovator absorbs and uses

it. If the innovator does not have the faculty to take this information in, it is likely that the

information may only play a minor role in her idea generation process (Cohen and Levinthal,

1990). If, however, the innovator is open to the experiences of her peer, then the information

can be better incorporated into the associational process used to generate ideas. Thus, increas-

ingly the likelihood that an individual generates a high-quality and novel idea. Conversely, even

if an innovator has all the traits that would allow her to absorb information and make novel

associations, if her conversational partner is unforthcoming, then relative to other more open

innovators, she will have less material to generate high-quality and novel ideas. The ideal match

therefore is a pairing between an extroverted peer—who willingly shares his experience—and

an innovator who shows behaviors related to greater openness. Thus, we hypothesize:

Hypothesis 2 High-openness individuals will develop higher quality ideas than those with less

openness after they talk to extroverted peers outside their team.

The prior theorized mechanisms are inherently directional in how they view the accumu-

10

Page 11: Sharique Hasan Rembrand Koning Stanford GSB …...Sharique Hasan Stanford GSB Rembrand Koning Stanford GSB September 29, 2015 Abstract Social interaction is thought to a ect individual

lated benefits of an extroverted–openness pairing. Extroverted individuals provide information,

open individuals absorb and use it to generate ideas. The converse may not hold: extroverted

individuals should not benefit as much from their conversations with open peers as as the open

peers benefit from talking to them. However, a key assumption underlying the directional

model is that the interaction is fundamentally a transfer and that little if anything “shared”

emerges through the interaction. An alternative perspective on the interaction dynamics in such

a pairing is that complementary matches affect not just the directional flow of information, but

qualitatively change a conversation’s character (McFarland, Jurafsky and Rawlings, 2013).

Consider a conversation between two innovators. Like before, an extroverted individual in-

teracts with one open to experience. In conversation, the former shares more. The latter is

curious and asks questions. This complementary back and forth in conversation is public such

that both parties can learn from the conversation and its constituent information. Specifically,

though the extroverted individual shares more of his experiences, he also benefits from hearing

and learning from the questions posed by her more open conversational partner. Furthermore,

the social ease that may emerge from such a pairing may create a positive feedback loop where

both parties eventually share and inquire—resulting in a high volume of information and expe-

rience accessible to both parties. In this scenario, both individuals benefit because the value of

the interaction is emergent and not just a directional transfer. Thus, we would hypothesize:

Hypothesis 3 Extroverted individuals will develop higher quality ideas than those who are less

extroverted after they talk to high-openness peers.

Extraversion, Openness and Team Dynamics

The proposed mechanisms have implications at the level of the team as well. Teams, rather

than lone individuals, constitute the more natural level of innovation in both markets and

organizations today (Guimera et al., 2005). Small teams, such as startups or internal product

teams rely on conversations with external peers at the ideation stage. External peers can

potentially provide teams with insight, information, perspective, problems, possible solutions,

and a host of other data. When individual team members interact with external peers, they

accumulate raw data that they can bring back to their team. A greater volume of accumulated

information should lead to wider and more high quality set of ideas generated by the team

(Reagans and Zuckerman, 2001; Reagans, Zuckerman and McEvily, 2004). However, the ability

11

Page 12: Sharique Hasan Rembrand Koning Stanford GSB …...Sharique Hasan Stanford GSB Rembrand Koning Stanford GSB September 29, 2015 Abstract Social interaction is thought to a ect individual

of a team to capitalize on the raw information it has accumulated depends on whether the

team creates novel associations and generate creative hypotheses from the information they

accumulate (Woolley et al., 2010; Girotra, Terwiesch and Ulrich, 2010).

Thus, a key facet of a team’s innovative process its ability to incorporate external information

and recombine it into new ideas (Paulus and Yang, 2000). Teams whose members interact with

extroverts should have more information; teams whose members are more open should be better

positioned to effectively recombine. Specifically, a team whose members interact with many

extroverts should be able to accumulate a large volume of information. This raw information

enhances the team’s ability to innovate in two ways. First, if individuals use only the information

gleaned from their own external conversations to generate ideas, teams with extrovert-open

pairings should have a higher quality pool of ideas from which to select. Thus, even if no cross-

sharing of external information across team members occurs, a team with higher levels of peer

extraversion and within-team openness should have higher quality ideas (Girotra, Terwiesch

and Ulrich, 2010). Second it provides a pool of shared data that the team can draw upon in

jointly generating ideas (Woolley et al., 2010). For instance, if individual A brings back novel

information from a peer conversation to the team, another team member B now has access to

this information and can build upon it to generate an idea. In this scenario, the team engages

in in a joint innovation process where the externally accumulated data is used by all members of

a team to develop ideas jointly. Regardless of which mechanism is operating, one should expect

the following to hold:

Hypothesis 4 Teams who converse with, on average, more extroverted peers outside their team

will develop higher quality ideas.

Hypothesis 5 Teams with higher levels of average openness will develop higher quality ideas

than those with less openness after they talk to extroverted peers outside the team.

Finally, not all teams should be equally placed to capture the returns to greater quantities

of information and higher levels of openness (Reagans, Zuckerman and McEvily, 2004; Woolley

et al., 2010). Internal team dynamics, particularly a team’s ability to have positive collaborative

experiences and coordinate effectively should affect whether it can develop novel ideas, even if

it has extroverted peers and open members. Specifically, the joint-innovation mechanism rests

upon the premise that team members effectively share information, allow for discourse, and

maintain an environment where team members feel comfortable generating novel ideas even if

12

Page 13: Sharique Hasan Rembrand Koning Stanford GSB …...Sharique Hasan Stanford GSB Rembrand Koning Stanford GSB September 29, 2015 Abstract Social interaction is thought to a ect individual

they are drawn from the extremities of the quality distribution (Woolley et al., 2010). Thus, we

hypothesize:

Hypothesis 6 Teams with high-quality collaboration dynamics will be better able to convert the

benefits of extroverted peers and team openness into higher quality ideas.

Empirical Setting and Methods

We propose that extroverted peers provide a greater volume of information to individuals; that

individuals with greater openness to experience benefit more from conversations with extroverted

peers; that teams with an extrovert-openness match generate better ideas; and that internal team

dynamics moderate the extent to which the team can capitalize on this process. Testing such a

theory is difficult for four important reasons. First, unlike formal collaborations, conversations,

especially with external peers are difficult to observe by researchers. They are rarely recorded

by those participating in them. Furthermore, conversations vary dramatically in their structure,

purpose, and length and comparing across conversations to test the impact of peer characteristics

is often confounded by these variations. Second, researchers rarely observe the dynamics of the

innovation process. What is often observed is the final product developed by a team and

rarely the raw ideas generated by individuals and team. Third, while researchers often have

data about peers’ level of education, publication records, or patents, more nuanced measures

of peer characteristics such as personality are generally difficult to gather, especially for peers

with whom an individual only has a conversation. Finally, and perhaps just as importantly,

the choice of both team and external conversational partners is endogenous—often self-selected

based on peer and individual characteristics that are unobservable to the researcher (Manski,

1993). Thus, the empirical demands for testing our hypotheses are considerable.

Experimental Design: An Innovation Competition

We overcome many of these hurdles by conducting a novel field experiment embedded in an

entrepreneurship bootcamp held in New Delhi, India in July of 20141. During the bootcamp,

112 aspiring entrepreneurs from across India participated in a three-week program intended to

1The experimental nature of the bootcamp was reviewed by our university’s Institutional Review Board. Allparticipants signed two consent forms, one online at the time of application and the second paper-based on the firstday of the bootcamp.

13

Page 14: Sharique Hasan Rembrand Koning Stanford GSB …...Sharique Hasan Stanford GSB Rembrand Koning Stanford GSB September 29, 2015 Abstract Social interaction is thought to a ect individual

help them develop skills in idea generation, design thinking, prototype design, and business

model development. The age range of the 112 graduates ran from 18 to 36, with a mean age

of just over 22. Our program had 25 women and everyone had, or was enrolled in, college

with 60 of the participants enrolled in a college, masters, or phd program. Our program was

regionally diverse with 62 of the participants from the state of Delhi and the rest from across

India. The class was primarily engineering and computer science degree holders (78), followed

by 18 business degrees, and the rest from the arts and sciences. Finally, 8 people were enrolled

in, or, had graduated from advanced degree programs.

The bootcamp provided instruction from leading member’s of India’s startup ecosystem in-

cluding successful entrepreneurs, designers, and venture capitalists. The program was structured

into three week-long modules. The first week, which was the most structured (and on which

we base this study) focused on the idea generation process. To incentivize participation and

effort, the three most highly rated proposals and prototypes from this week won prizes totaling

45,000 Indian Rupees (789.47 dollars). The major prizes were team based. The first prize was

20,000 INR, second 10,000 INR, and third 7,500 INR. The prize allocation was based on the

average rating a team’s proposal received during the peer review process, with the three highest

teams winning the top three prizes. The second week focused on business models; the final week

was the least structured and participants could select their own teams of three people from the

bootcamp to develop a business concept and prototype to receive up to 8,000 USD in funding

and support to implement their idea.

We use the activities from the first week and data collected before the bootcamp began to test

our hypotheses. Before the bootcamp began, we asked all participants to complete surveys, chief

among these was the 44-item Big Five Inventory (John and Srivastava, 1999). All participants

allowed us to collect pre-bootcamp (and thus, pretreatment) measures of Extraversion, Openness

to Experience, Neuroticism, Agreeableness and Conscientiousness. We discuss the construction

of our key variables using this inventory in the variables section below.

The first day (Monday) was dedicated to logistics, an introduction to the program, and a

short icebreaker in a randomized group at the end of the day. We did not collect any data during

this day, as it was not part of the experimental setup for the week. The second day (Tuesday)

began with individuals reporting to one of forty tables where they sat with their icebreaker

group and were asked to individually generate as many or as few ideas for innovative software

products for the Indian wedding industry. The text of the prompt read:

14

Page 15: Sharique Hasan Rembrand Koning Stanford GSB …...Sharique Hasan Stanford GSB Rembrand Koning Stanford GSB September 29, 2015 Abstract Social interaction is thought to a ect individual

On November 27, 2011, over 60,000 weddings took place on this one single day in New

Delhi, just because the day was auspicious. Every wedding hall in Delhi was booked

for every shift and families paid large premiums of at least 1 to 2 lakhs, or more to

book even the smallest halls. Even on less auspicious days, Indian weddings are big,

fun, complex, loud, colourful, and most of all expensive. Today, the size of the Indian

wedding industry is estimated to be around 2.25 trillion Indian rupees or 38 billion

US dollars. The industry is also diverse—it includes such products and services as

marriage gardens, match-making, clothing, decorations, makeup, gifts, jewelry, and a

lot more. Startups in India have only scratched the surface of this industry—the most

prominent example is Shaadi.com which has revolutionized matchmaking and made

many aunties across India obsolete. Your task for this week is to develop a product

concept for a mobile and web app that will reinvent part of the wedding experience–

either before the wedding, during the wedding and after the wedding—in India. On

to reinventing!

We chose the Indian wedding industry as our prompt for three reasons. First, based on

conversations with Indian entrepreneurs and venture capitalists, the Indian wedding industry is

large and has significant market potential. Several venture capital firms are actively investing in

software products for this large market. The choice of the wedding industry was therefore based

in part on concerns of external validity. Second, unlike finance or biotechnology, the “Indian

wedding” was something that that the vast majority of bootcamp participants had experienced,

but was an industry where a subset of individuals would not have a systematic skill or knowledge

advantage. Third, we chose this industry because it was a relatively diverse domain, constituting

problems ranging from finding mates to buying wedding dresses to honey-moon selection and

even post-marital counseling. Thus, the Indian wedding context had the potential to produce

a differentiation in the types and quality of ideas generated by the participants. For one hour,

the participants entered each discrete idea into a software application as short texts. Figure

1 depicts the prompt and entry screen for a participant. Individuals generated 6.6 ideas with

each idea having a length of approximately 505 characters. We call these ideas “pre-treatment”

ideas.

[Figure 1 about here.]

Conversational Peer Randomization. To test our hypotheses, we randomized a set of three

15

Page 16: Sharique Hasan Rembrand Koning Stanford GSB …...Sharique Hasan Stanford GSB Rembrand Koning Stanford GSB September 29, 2015 Abstract Social interaction is thought to a ect individual

empathy interviews that participants had with other members of the bootcamp. These inter-

views are a staple of the Design Thinking approach (Kelley and Kelley, 2013). Each empathy

interview lasted 14 minutes and consisted of a random pairing between two individuals at the

bootcamp. We put each pair in random and preassigned seats across from each other, with

participants assigned (randomly) to an “A” and a “B” position. The protocol of the interview

was semi-structured and participants were asked to learn about their conversational peer’s expe-

rience with an Indian wedding. We began with person “A” interviewing and listening to person

“B’s” perspective for four minutes; followed by person “B” interviewing and listening to person

“A’s” perspective for the same amount of time. Next, person “A” was asked to “dig deeper”

by asking person “B” more questions for 3 more minutes; followed by person “B” repeating this

process with person “A.” During and after the conversation, the participants could take notes

about their conversation and could record it in the sheet depicted in Figure 2. After the first

pairwise peer interaction, individuals were re-randomized to two more pairwise interactions fol-

lowing the same structure. After all three randomizations, individuals were instructed to return

to their original assigned table and generate new ideas for one hour. Participants generated an

average of 4.5 ideas with the average idea having 476 characters. We call these “post-treatment”

ideas.

[Figure 2 about here.]

Anonymous Peer Evaluations of Individual Ideas. The next morning from 9.30am to 11am

(Wednesday; Day 3), all participants anonymously evaluated a random subset of both the pre-

and post-treatment ideas of other bootcamp participants. Our choice of double-blind anonymous

peer evaluations arises from three considerations. First, peer evaluation is perhaps the most

common evaluation in many contexts. In academia, research articles are evaluated anonymous

peers as are grants (Marsh, Jayasinghe and Bond, 2008); In organizations, many decisions

about products and design choices are evaluated by peers; In education, peer evaluations are

becoming increasingly common for evaluating classroom projects (Cooper and Sahami, 2013;

Reily, Finnerty and Terveen, 2009). Second, many prior studies of creativy have used peer

ratings as measures of the creative output of teams and individuals (Amabile et al., 2005, 2004).

Third, peer evaluation, particularly in this context is superior to evaluations by external or online

parties who may not have either the incentive or ability to effectively assess an idea’s worth.

Finally, research indicates that peer evaluations are more accurate when evaluators are blinded

16

Page 17: Sharique Hasan Rembrand Koning Stanford GSB …...Sharique Hasan Stanford GSB Rembrand Koning Stanford GSB September 29, 2015 Abstract Social interaction is thought to a ect individual

to the identity of the subject. They are also harsher and more accurate when evaluating more

than three items (Marsh, Jayasinghe and Bond, 2008; Boudreau et al., 2012). Thus, we asked

individuals to rate approximately 50 ideas; on three dimensions on a 5-point likert scale strongly

disagree to strongly agree: whether the idea was novel; whether the product was something

that the rater would buy, and whether the idea had business potential. Each idea received

approximately 6.24 ratings. The average rating of business value was 2.45; buy likelihood of

2.59; and of novelty 2.43.

Idea development in teams. At the end of the evaluation session on day 3, individuals were

randomly assigned teams of approximately three individuals. It is within these teems that

individuals worked on day 3, day 4 and day 5 to develop a mock-up prototype and business

plan. The team was given the freedom to work on any idea that they jointly chose. The idea

could be one from the pre-treatment ideation session, the post-treatment session, a combination

of both, or neither. By midnight of day five (Friday) participants submitted a complete packet

of the prototype which included a “splash page” consisting of a graphic describing their product;

a presentation walk-through of their software prototype; a text description of their product and

the problem it was intended to solve; a one sentence description of their product; and a product

name.

Final submission evaluations and rating of team dynamics On day six (Saturday) we assigned

the 112 participants five random and anonymous project submissions to to evaluate (excluding

their own). Participants evaluated their assigned submissions using an online system where

students ranked projects from the best to the worse in the subset of five assigned submissions.We

recoded the rankings so that a ranking of 5 indicated the best submission and a ranking of 1

indicated the worst. Each submission therefore received approximately 14 evaluations on 12

dimensions including: product novelty, unique insight, display of empathy for customer needs,

feasibility, business potential, as well as the quality of the prototype walkthrough and splash page

(Girotra, Terwiesch and Ulrich, 2010). Furthermore, team members also rated their teammates

using the Society for Human Resource Management’s 24-item “Manager Effectiveness Scale”

360 degree evaluation. These questions concerned individual team members’ pro-collaborative

behaviors within the team during the first week.

Figure 3 summarizes the process of the experiment, the randomizations, and data collection.

[Figure 3 about here.]

17

Page 18: Sharique Hasan Rembrand Koning Stanford GSB …...Sharique Hasan Stanford GSB Rembrand Koning Stanford GSB September 29, 2015 Abstract Social interaction is thought to a ect individual

Individual-level Hypotheses

Dependent variables. Our first set of hypotheses concerns the relationship between conversa-

tional peer extraversion on the quality of post-treatment ideas generated by individual innovators.

The key dependent variables for this analysis derive from the anonymous peer evaluations (day

three) of the raw ideas generated by individuals on day two. The first of these dependent vari-

ables is Idea Quality and it is the sum of the evaluations an idea receives from an anonymous

evaluator on the dimensions of business value, buy likelihood, and novelty.2 We also model

the relationship between our key independent variables on the three individual ratings as well:

Business, Buy, and Novelty.

Independent variables To examine the relationship between peer extraversion and idea qual-

ity, we create a variable for the average extraversion score of an individual’s three randomly

assigned conversational peers. Extraversion scores are calculated average of the positively coded

likert scores for the 8-item extraversion scale from the BFI. We call this variable Extraversion

(peer). We also create a variable called Openness (self) measuring the average of an individual’s

positively coded responses from the 10-item openness scale. Finally, we created an interaction

variable called Extraversion (Peer) × Openness (Self) measuring the product of conversational

peers’ extraversion scores and the focal individual’s own openness score. All independent vari-

ables are rescaled so that means are “0” and standard deviations are “1.”

Causality and control variables. The key strength of our empirical analysis is the experi-

mental design, which consists of randomization and lagged independent variables. The lagged

covariates and randomization allow us to deal with two of the main inferential challenges in the

study of social influence: the reflection problem and the selection problem. Lagging our peer

extraversion measure ensures that a focal individuals’ post-treatment idea quality does not cause

increases in her peers’ level of extraversion. Further, the random assignment of conversational

peers reduces the likelihood that peer interaction emerges from self-selection and homophily

(Manski, 1993, 2000). However, scholars recommend controlling for other covariates even in

research designs with randomization (Imai, King and Stuart, 2008). Thus we include in several

of our models, controls for a focal individual’s own extraversion, as well as scores for both the

focal individual and peers’ level of agreeableness, neuroticism and conscientiousness. Further-

more, we include a variable called Ability for both the focal individual and peers’ measuring the

2Some ideas received evaluations on all dimensions while others received evaluations on only one. For the con-struction of Idea Quality we code as missing this score if it does not receive evaluations on all three dimensions.

18

Page 19: Sharique Hasan Rembrand Koning Stanford GSB …...Sharique Hasan Stanford GSB Rembrand Koning Stanford GSB September 29, 2015 Abstract Social interaction is thought to a ect individual

average rating of each person’s application received from four independent evaluators during the

admission process. The application scores were reverse coded so that the better applications

received high scores and worse applications received lower scores. Raters assessed ability based

on the participants grades in college, the prestige of their college, the quality of their application

essay, as well as their skills in business topics such as finance, marketing and sales as well as

technical skills such interaction design and programming. Table 1 presents summary statistics

for our key variables. Furthermore, we correlated measures of our key treatment Extraversion

(peer) with pretreatment characteristics of a focal individuals’ idea quality and traits and find,

as expected, that they are uncorrelated. Table 2 presents these bivariate regressions.

[Table 1 about here.]

[Table 2 about here.]

To test our key hypotheses we used ordered logistic regression models to regress all eval-

uations e of idea d by individual i on the her randomized conversational peers’ average level

of extraversion. Further, we include both the main effect of individual i’s openness and the

interaction between peer extraversion and individual openness to test the match hypotheses.

Since we have multiple evaluations and multiple ideas for individuals i, we correct our standard

errors by clustering them at the individual level, thus creating 112 clusters for the individuals

participating in the study3.

Team-level Hypotheses

Dependent variables. Our second set of hypotheses concerns how conversations with extroverted

peers affects team performance and how individual openness at the level of the team moderates

this effect. To conduct this analysis we construct four dependent variables. As mentioned

earlier, at the week’s end (Saturday, day 6) individuals conducted double-blinded evaluations of

five projects randomly selected from the 39 other submissions (excluding one’s own submission).

Using the reverse coded ranking along the 12 dimensions (5 best, 1 worst) we created four indices.

The first is Total which measures the average of the rankings a project received across all 12

dimensions. Higher scores mean better rated projects. Furthermore, we created a variable called

Novelty and Insight which was the average of the Novelty, Insight and Empathy rankings. To

3Some individuals did not generate ideas at this stage or their ideas were not evaluated on all dimensions by theevaluators, thus in some models we have fewer than 112 clusters

19

Page 20: Sharique Hasan Rembrand Koning Stanford GSB …...Sharique Hasan Stanford GSB Rembrand Koning Stanford GSB September 29, 2015 Abstract Social interaction is thought to a ect individual

measure a submission’s business potential we created a variable called Business that measures

the average of the proposed product’s feasibility and business potential. Finally, we created a

variable called Prototype measuring the average of the rankings a submission received for its

visual splash page and it’s prototype walk through slides.

Independent variables. To test our hypotheses regarding extroverted peers, average team

openness, and team dynamics we created five variables. The first variable Extroverted Peers

(team) that measures the average of all team members’ peer extraversion scores: we calculate

the average score within a team of every team member’s Extraversion (peer) score. The second

variable we construct is Openness (team) that measures the average level of a team’s openness

to experience. To create the interaction variable of the level of a team’s peers’ extraversion and a

team’s level of openness, we interact the latter two variables to create Extroverted Peers (team)

× Openness (team). Finally, we create a variable called Team Dynamics measures the mean

of each individual’s average score on the 360 feedback “Manager Effectiveness Scale.” Again,

we normalize all independent variables to have a mean of “0” and a standard deviation of “1.”

Table 3 presents summary statistics for our key variables. Furthermore, we correlate measures

of our key treatment Extraversion (peer) with pretreatment characteristics of a focal team and

find, as expected due to our randomization, that they are uncorrelated. Table 4 presents these

bivariate regressions.

To test our key hypotheses we used linear regression models to regress all evaluations e of

submission s by team i on the team’s randomized conversational peers’ average level of extraver-

sion. Further, we include both the main effect of team i’s average openness and the interaction

between peer extraversion and team openness to test the match hypotheses. Further, to test

our hypotheses that within-team dynamics moderates whether information advantage arising

from the extraversion-openness match affects team performance, we interact our independent

variables with the Team Dynamics score. Since we have multiple evaluations for each team’s

ideas, we correct our standard errors by clustering them at the team level, thus creating 40

clusters, one for each team.

[Table 3 about here.]

[Table 4 about here.]

20

Page 21: Sharique Hasan Rembrand Koning Stanford GSB …...Sharique Hasan Stanford GSB Rembrand Koning Stanford GSB September 29, 2015 Abstract Social interaction is thought to a ect individual

Results

Do conversations with extroverted peers increase idea quality?

We begin our analysis by examining whether individuals develop better rated ideas if they have

extroverted conversational peers (Hypothesis 1). We regress measures of idea quality on the

average extraversion of an individuals’ randomly assigned peers. Table 5 presents the first set

of estimations. All models cluster correct standard errors at the individual i level to account

for multiple evaluations of an individual’s ideas.

Column 1 presents estimates of peer extraversion on the aggregate post-treatment Idea Qual-

ity measure. The coefficient is both positive and statistically significant (p < .01), suggesting

that when individuals have conversations with extroverted peers, they generate better rated

ideas. The magnitude of this effect can be more easily interpreted as log-odds, by exponen-

tiating the coefficient we find that the log odds for the peer extraversion variable are 1.165.

This suggests that for individuals who have extroverted peers one unit higher than average, are

about 16% more likely to receive ratings higher than those individuals with the mean level of

extraversion. To test the robustness of this result, in Column 2 we include a measure of a focal

individual’s own extraversion and find that it does not alter our coefficient estimate substan-

tively. To ensure that our measure of extraversion is not related to some measure of a peer’s

ability or education, we include measures of a peers’ average Ability. While the coefficient is

positive, it is small and statistically indistinguishable from zero. Similarly, in column 4, we

include measures of the focal individual’s ability as well. Inclusion of this variable does not

alter our effect. In columns 5 and 6, we present a complete model and also include a measure

of the average quality of a focal individual’s pre-treatment ideas. We find a modest effect of

pre-treatment idea quality on the post-treatment idea quality. However, our main effect of peers’

extraversion remains stable. Thus, we find strong support for hypothesis 1.

[Table 5 about here.]

In table 6 we examine whether our results hold in the more disaggregated versions of the

post-treatment idea ratings. We find that indeed, peer extraversion increases the quality of

ideas on the dimensions of business value, buy likelihood, and novelty. Further, our results also

indicate that if an individual developed novel pre-treatment ideas, they also developed novel

post-treatment ideas. For comparison, the estimate of pre-treatment idea novelty is approxi-

21

Page 22: Sharique Hasan Rembrand Koning Stanford GSB …...Sharique Hasan Stanford GSB Rembrand Koning Stanford GSB September 29, 2015 Abstract Social interaction is thought to a ect individual

mately the same size as that of peer extraversion. Suggesting the importance of greater volumes

of external information in the idea generation process.

[Table 6 about here.]

Does a focal actor’s openness moderate the effect of extroverted peers?

In tables 7 and 8 we examine the hypothesis that an individual’s openness to experience mod-

erates the extent to which peer extraversion affect idea quality (Hypothesis 2). To do so, we

re-estimate the models above and a variable for an individual’s level of openness and an in-

teraction of this variable with her peers’ average extraversion. Columns 1 through 3 in table

7 show that when the focal individual has a one unit higher level of openness and her peers

are extroverted, her ideas are rated higher than if she had extroverted peers alone. Regarding

magnitude, we find that the log-odds effects of such a match on an ideas business value increase

from1.25 to 1.36, buy likelihood increases from 1.25 to 1.47, and novelty increases from 1.18 to

1.24. Thus, we find a substantial benefit of openness for incorporating the information trans-

mitted by extroverted peers. In table 8 we test for this effect with the aggregated score. In

column 1, we find that a one-unit increase in openness increases the log-odds of a higher rating

from 1.32 to 1.40. Thus, we find support for hypothesis 2 that openness increases the value of

interaction with extroverted peers.

[Table 7 about here.]

[Table 8 about here.]

To test hypothesis 3, that the benefits to an extraversion-openness match are shared by both

parties, we examine whether benefit exists for extroverted individuals when they are paired with

peers with greater openness to experience. Tables 7 and 8 present these results. Our results here

suggest that such a match does not benefit the focal individual. Their ideas are, if anything, less

good or no better when they talk to more open people. We find little support for hypothesis 3.

To ensure that our measures of peer extraversion and focal individual openness are not

picking up on other personality traits, we included controls for other peer traits and interaction

with own traits. Our results hold under these specifications, giving us greater confidence about

our posited mechanisms. These results are presented in table 9.

[Table 9 about here.]

22

Page 23: Sharique Hasan Rembrand Koning Stanford GSB …...Sharique Hasan Stanford GSB Rembrand Koning Stanford GSB September 29, 2015 Abstract Social interaction is thought to a ect individual

We also wanted to ensure that our estimates were not biased by whether an evaluator had

knowledge of or interacted with the individual generating the idea they were evaluating anony-

mously. To do this, we conduct an analysis where we control for the presence of a relationship

prior to the treatment between an evaluator and the focal innovator. Our results, presented in

table 10 indicate that such a bias does not appear to exist nor does it affect our key results.

Further, our results hold if we drop all evaluations conducted by an evaluator who knew the

individual whose blinded idea they were evaluating. Again, our results remain robust.

[Table 10 about here.]

These results together provide strong evidence that peer interaction—namely interaction

with extroverted peers increases the quality of individual ideas. Further, we find evidence that

this effect is larger for individuals more open to experience.

Peers, Teams and Innovation

Our second set of hypotheses concern the relationship between peer extraversion, team openness

and the dynamics of the collective innovation within a team. To test these hypotheses we regress

measures of final submission quality on the team-level measures of peer extraversion, team

member openness, and team member collaborative skill. Table 11 presents the estimations for

this analysis. All models cluster correct standard errors at the team level to account for multiple

evaluations of a team’s submissions.

Column 1 regresses total final project score Total on team level measures of our key inde-

pendent variables. Our results provide modest support for hypothesis 4 that the benefits of

extroverted peers also extend to the final ideas generated by teams. We find that the main

effect of peer extraversion is positive and statistical significant at the p = .057 level when we

look at total score and becomes large and more significant for the novelty model in Column 2.

Further, we find that the interaction of team openness and peer extraversion are also positive

and statistically significant at least at the p < .1 level for the total rating, novelty, and business

value models. This suggests that teams with more open members are better able to generate a

pool of ideas with which to generate one high quality product concept. These results also pro-

vide support for hypothesis 5, that an extraversion-openness match also increases idea quality

at the team level.

23

Page 24: Sharique Hasan Rembrand Koning Stanford GSB …...Sharique Hasan Stanford GSB Rembrand Koning Stanford GSB September 29, 2015 Abstract Social interaction is thought to a ect individual

Finally, we test hypothesis 6 which argues that when internal team dynamics are cohesive—

that individuals seek input from others, show concern, seek diverse feedback, encourage diverse

perspectives, and other integrative and cohesive behaviors—the team is better able to incorpo-

rate external ideas in their innovation process. Column 5 tests this hypothesis by including the

main effect of team member cohesiveness as well as an interaction with the extraversion-openness

match variable. We find that although the main effect of the positive team dynamics variable is

not statistically significant, the interaction is significant and of substantial magnitude (β = .493,

p < .05). Thus, our results indicate strong support for the idea that more cohesive behaviors by

team members allow the team to better incorporate and capitalize on the acquired information

and generated raw ideas. Finally, it is worth noting that high ability teams developed higher

rated prototypes than teams with lower-ability team members.

[Table 11 about here.]

Discussion

Social networks and social interaction have become central components of the standard socio-

logical model of innovation (Burt, 2004; Fleming, Mingo and Chen, 2007; Reagans, Zuckerman

and McEvily, 2004; Obstfeld, 2005). In this article, we incorporate the role of social psycholog-

ical processes into a more structural model of peer interaction to show how individual traits,

interpersonal dynamics and team dynamics moderate how purely structural variance in peer

interaction affects the generation of high-quality and novel ideas. Our key finding is that indi-

viduals generate better ideas when they have conversations with more talkative conversational

partners willing to share their experiences—i.e. extroverted peers. The magnitude of this effect

is modest, suggesting that there a one unit increase in peer extraversion increases the log-odds

that an idea is rated higher by 16% or approximately .25 points on a 5 point scale. While this

effect will not alone make the lowest ability quality ideas the best ones, it can shift ideas at the

margins of “good.” to become “very good.” Further, our experimental treatment in this article

only varied peer interaction for three, 20-minute sessions. It is likely that the magnitude of this

effect will be greater if peer interaction were more frequent and had greater duration.

Our findings, however, also dovetail with a growing body of recent research on the contin-

gent effects of social interaction on a host of outcomes. For extroverted peers to be most useful

during the innovation process, innovators should be open to experience to benefit from the high

24

Page 25: Sharique Hasan Rembrand Koning Stanford GSB …...Sharique Hasan Stanford GSB Rembrand Koning Stanford GSB September 29, 2015 Abstract Social interaction is thought to a ect individual

volume of information provided by their extroverted conversational peers. The magnitude the

extraversion–openness match results indicate that individuals who have one unit more extro-

verted peers and are themselves one unit more open than the average match, have ideas rated

higher by 25% to 40%, depending on the metric. These results strongly indicate that social

interaction should not be viewed as a unilateral transfer of information, with receivers as pliant

acquirers of external resources. Indeed, our findings indicate that variance in individual traits

shapes the degree to which individuals can capitalize on their social interactions.

A key benefit of our study is our ability to trace how early peer conversations translate into

team performance. We argue that at the level of the team the aggregation of information from

extroverted peers combined with team members’ own openness affects the ability of a team to

develop an idea. Our findings support this claim. We find evidence for both the main effect

of peer extraversion at the aggregate level of the team and evidence that this effect increases

when an extraversion–openness match exists. This finding gives us greater confidence in the

individual level results and also suggests that the knowledge generated by individuals through

peer interaction is not just an individual resource, but rather a resource shared by the team.

Finally, we find that some teams are better able to use and recombine the information

acquired by their individual team members. Teams that have more collaborative members—

i.e. those that share ideas, listen to feedback, and allow for discourse—and also have high

extraversion-openness matches have better rated ideas. To provide a sense of magnitude, a one-

unit increase on peer extraversion, team member openness and team member cohesion increases

the average rating of an idea by .488 points out of 5. While the size of these combined effects

indicates that that while such access to external information and good team dynamics alone

will not take a bad idea and make it great ; it could take a good idea and make it very good or

make a very good idea great.

Conclusions

This study holds several implications for the study of innovation, particularly the role that social

interaction plays in the generation of novel ideas both at the level of the team and the individual.

The primary contribution of this article is the embedding of social-psychological processes in a

structural model of innovation. First, we build on the work of Reagans, Zuckerman and McEvily

(2004), as well as network and innovation scholars such as Burt (2004) and Fleming, Mingo and

25

Page 26: Sharique Hasan Rembrand Koning Stanford GSB …...Sharique Hasan Stanford GSB Rembrand Koning Stanford GSB September 29, 2015 Abstract Social interaction is thought to a ect individual

Chen (2007) by incorporating the role that concrete conversations and social interaction have

in the information acquisition process from network ties. Our results support the idea that

external networks provide important information to individuals and teams. Furthermore, our

results indicate the criticality of matches in focal-individual and peer traits for information flow

to be most beneficial. These results provide a social psychological basis for individual level

differences in absorptive capacity. Third, and finally, our results highlight the importance of

intra-team processes—namely cohesion—in increasing the ability for teams to capitalize on this

flow.

Our results also have implications beyond the context of innovation. Scholars in a variety of

domains have been studying the role of external peers and social networks in creating divergences

in important outcomes. Perhaps one of the largest of these literatures is the study of peers and

peer groups in the production of human capital (Sacerdote, 2001). While research has found

mixed evidence for the existence of peer effects in this context (Sacerdote, 2014), recent work

is beginning to suggest that matching processes, like the one described here, moderate the

extent to which learning happens across peers in educational contexts (Carrell, Sacerdote and

West, 2013; Hasan and Bagde, 2013). We also think the general framework of conceptualizing

peer effects as matching processes should also apply to network processes within labor markets

(Fernandez and Fernandez-Mateo, 2006), organizational promotion contexts (Burt, 1992), and

other domains where social interaction affects the flow of knowledge and information. Further,

our results lend support to the idea that the design of social systems to improve individual and

group level outcomes is not merely a process of matching people with resources to those without

them. Interpersonal dynamics and other kinds matching must be incorporated into the design

of matches.

Our study contributes to the literature from a methodological perspective as well. In this

article, we use data a from a field setting—and entrepreneurship bootcamp—in which we em-

bedded a randomized field experiment. By randomizing social interaction, namely external peer

conversations and team assignments, as well as measuring detailed data ideation and individ-

ual characteristics we could trace a nuanced and dynamic process from its earliest inception (a

conversation with an external peer) to the performance of teams five days afterwards. While

the specific context of our study is clearly not generalizable to all contexts, we imagine that

longer-duration field experiments with a subset of the innovations introduced in our study could

be used to, for instance study how external peer conversation affect the success of startups in

26

Page 27: Sharique Hasan Rembrand Koning Stanford GSB …...Sharique Hasan Stanford GSB Rembrand Koning Stanford GSB September 29, 2015 Abstract Social interaction is thought to a ect individual

more naturalistic contexts. This can be accomplished, for example, by working with incuba-

tors to facilitate external conversations between startups and an external pool of entrepreneurs,

venture capitalists, and customers.

In conclusion, we note several limitations of the present study. As with many field experi-

ments, our findings though benefiting from randomized peer interaction and detailed measure-

ment relies on a very specific context—an entrepreneurship bootcamp held in New Delhi, India.

Thus, our findings may not have broad generalizability outside of the early entrepreneurial

context. However, we think that future research, by accumulating a greater body of findings

across a wide array of contexts has the potential to sharpen our understanding of these network

processes. Further, while our empirical tests are conducted in one specific context, our theo-

retical model is build from general assumptions derived from a large body of literature both in

sociology, psychology and management (Fleming, Mingo and Chen, 2007; Reagans, Zuckerman

and McEvily, 2004; Reagans and Zuckerman, 2001; Cummings, 2004; Paulus and Yang, 2000;

Woolley et al., 2010).

Moving forward, we think several opportunities exist for extending our results and adding

even further nuance. One key limitation we have in this study is that the conversations between

the focal and the peer are never observed. Thus, we must infer based on our understanding

of the structure of the interaction and psychological theory the nature and volume of informa-

tion transmitted. However, we may learn much more about what the actual benefits are of

such extraversion-openness matches. Similarly, other types of measurement into interpersonal

interaction is likely to give greater depth to our theories. Another possibility for extending the

work in this paper is to take the idea of matching and formally incorporate it into our empirical

tests. In this article, we used randomization to create variance in peer and focal individual

characteristics within a set of pairings. Matches, thus, are a byproduct of the randomization

process and are not explicitly designed. Future designs, particularly for match effects decide in

randomized contexts should be designed to see whether they hold. Recent work by Carrell, Sac-

erdote and West (2013) suggests that implementing policy based on findings from randomized

studies may not necessarily yield expected results due to interpersonal dynamics. Finally, we

think an important future direction for this research is studying the implications of our findings

at a macro scale: more connections between individuals and organizations increase the overall

innovative capabilities of ecosystems and regions (Saxenian, 1996). The ideal situation would

be that network ties such as the ones created here increase performance for all members, and

27

Page 28: Sharique Hasan Rembrand Koning Stanford GSB …...Sharique Hasan Stanford GSB Rembrand Koning Stanford GSB September 29, 2015 Abstract Social interaction is thought to a ect individual

not just reshuffle the outcome distribution.

28

Page 29: Sharique Hasan Rembrand Koning Stanford GSB …...Sharique Hasan Stanford GSB Rembrand Koning Stanford GSB September 29, 2015 Abstract Social interaction is thought to a ect individual

References

Amabile, Teresa M, Elizabeth A Schatzel, Giovanni B Moneta and Steven J Kramer. 2004. “Leaderbehaviors and the work environment for creativity: Perceived leader support.” The Leadership Quar-terly 15(1):5–32.

Amabile, Teresa M, Sigal G Barsade, Jennifer S Mueller and Barry M Staw. 2005. “Affect and creativityat work.” Administrative science quarterly 50(3):367–403.

Aral, Sinan and Marshall Van Alstyne. 2011. “The diversity-bandwidth trade-off1.” American Journalof Sociology 117(1):90–171.

Azoulay, Pierre, Joshua Graff Zivin and Jialan Wang. 2010. “Superstar extinction.” Quarterly JournalOf Economics 25:549–589.

Beshears, John, James J Choi, David Laibson, Brigitte C Madrian and Katherine L Milkman. 2015.“The effect of providing peer information on retirement savings decisions.” The Journal of Finance70(3):1161–1201.

Blank, Steve. 2013. “Why the lean start-up changes everything.” Harvard Business Review 91(5):63–72.

Borgatti, S. P. and R. Cross. 2003. “A relational view of information seeking and learning in socialnetworks.” Management Science 49(4):432–445.

Boudreau, Kevin J, Eva C Guinan, Karim R Lakhani and Christoph Riedl. 2012. “The novelty paradoxand bias for normal science: evidence from randomized medical grant proposal evaluations.” HarvardBusiness School Technology & Operations Mgt. Unit Working Papers, SSRN .

Brebner, John and Chris Cooper. 1978. “Stimulus-or response-induced excitation. A comparison of thebehavior of introverts and extraverts.” Journal of Research in Personality 12(3):306–311.

Brown, Tim et al. 2008. “Design thinking.” Harvard business review 86(6):84.

Burt, R. S. 1992. Structural Holes: The Social Structure of Competition. Harvard University Press.

Burt, R. S. 2004. “Structural holes and good ideas.” American Journal of Sociology 110(2):349–399.

Carrell, Scott E, Bruce I Sacerdote and James E West. 2013. “From natural variation to optimal policy?The importance of endogenous peer group formation.” Econometrica 81(3):855–882.

Cheng-Hua, Wang, Lee Yuan-Duen, Lin Wei, Zhuo Li-Ting et al. 2007. “Effects of personal qualitiesand team processes on willingness to share knowledge: An empirical study.” International Journalof Management 24(2):250.

Cohen, Wesley M and Daniel A Levinthal. 1990. “Absorptive capacity: a new perspective on learningand innovation.” Administrative science quarterly pp. 128–152.

Cooper, Steve and Mehran Sahami. 2013. “Reflections on Stanford’s MOOCs.” Communications of theACM 56(2):28–30.

Csikszentmihalyi, Mihaly. 1997. “Flow and the Psychology of Discovery and Invention.” HarperPeren-nial, New York 39.

Cummings, Jonathon N. 2004. “Work groups, structural diversity, and knowledge sharing in a globalorganization.” Management science 50(3):352–364.

De Vries, Reinout E, Bart Van den Hooff and Jan A de Ridder. 2006. “Explaining knowledge sharingthe role of team communication styles, job satisfaction, and performance beliefs.” CommunicationResearch 33(2):115–135.

29

Page 30: Sharique Hasan Rembrand Koning Stanford GSB …...Sharique Hasan Stanford GSB Rembrand Koning Stanford GSB September 29, 2015 Abstract Social interaction is thought to a ect individual

Diette, Timothy M and Ruth Uwaifo Oyelere. 2014. “Gender and Race Heterogeneity: The Impact ofStudents with Limited English on Native Students’ Performance.” The American Economic Review104(5):412–417.

Fernandez, Roberto M. and Isabel Fernandez-Mateo. 2006. “Networks, Race, and Hiring.” AmericanSociological Review 71(1):42–71.

Fleming, Lee, Santiago Mingo and David Chen. 2007. “Collaborative brokerage, generative creativity,and creative success.” Administrative Science Quarterly 52(3):443–475.

George, Jennifer M and Jing Zhou. 2001. “When openness to experience and conscientiousness arerelated to creative behavior: an interactional approach.” Journal of applied psychology 86(3):513.

Girotra, Karan, Christian Terwiesch and Karl T Ulrich. 2010. “Idea generation and the quality of thebest idea.” Management Science 56(4):591–605.

Goh, Swee C. 2002. “Managing effective knowledge transfer: an integrative framework and some practiceimplications.” Journal of knowledge management 6(1):23–30.

Grant, Adam M and James W Berry. 2011. “The necessity of others is the mother of invention:Intrinsic and prosocial motivations, perspective taking, and creativity.” Academy of ManagementJournal 54(1):73–96.

Guimera, Roger, Brian Uzzi, Jarrett Spiro and Luis A Nunes Amaral. 2005. “Team assembly mechanismsdetermine collaboration network structure and team performance.” Science 308(5722):697–702.

Hartmann, Wesley R, Puneet Manchanda, Harikesh Nair, Matthew Bothner, Peter Dodds, David Godes,Kartik Hosanagar and Catherine Tucker. 2008. “Modeling social interactions: Identification, empir-ical methods and policy implications.” Marketing letters 19(3-4):287–304.

Hasan, Sharique and Surendrakumar Bagde. 2013. “The Mechanics of Social Capital and AcademicPerformance in an Indian College.” American Sociological Review 78(6):1009–1032.

Hasan, Sharique and Surendrakumar Bagde. 2015. “Peers and Network Growth: Evidence from aNatural Experiment.” Management Science .

Imai, Kosuke, Gary King and Elizabeth A Stuart. 2008. “Misunderstandings between experimental-ists and observationalists about causal inference.” Journal of the royal statistical society: series A(statistics in society) 171(2):481–502.

John, Oliver P and Sanjay Srivastava. 1999. “The Big Five trait taxonomy: History, measurement, andtheoretical perspectives.” Handbook of personality: Theory and research 2(1999):102–138.

Kelley, Tom and David Kelley. 2013. Creative confidence: Unleashing the creative potential within usall. Crown Business.

Lee, Choonwoo, Kyungmook Lee and Johannes M Pennings. 2001. “Internal capabilities, externalnetworks, and performance: A study on technology-based ventures.” Strategic management journal22(6-7):615–640.

Lerner, Josh and Ulrike Malmendier. 2013. “With a little help from my (random) friends: Success andfailure in post-business school entrepreneurship.” Review of Financial Studies p. hht024.

Liao, Shu-Hsien, Wu-Chen Fei and Chih-Chiang Chen. 2007. “Knowledge sharing, absorptive capacity,and innovation capability: an empirical study of Taiwan’s knowledge-intensive industries.” Journalof Information Science 33(3):340–359.

Manski, C.F. 1993. “Identification of endogenous social effects: The reflection problem.” The Reviewof Economic Studies 60(3):531.

30

Page 31: Sharique Hasan Rembrand Koning Stanford GSB …...Sharique Hasan Stanford GSB Rembrand Koning Stanford GSB September 29, 2015 Abstract Social interaction is thought to a ect individual

Manski, C.F. 2000. “Economic Analysis of Social Interactions.” The Journal of Economic Perspectives14(3):115–136.

Marsh, Herbert W, Upali W Jayasinghe and Nigel W Bond. 2008. “Improving the peer-review process forgrant applications: reliability, validity, bias, and generalizability.” American Psychologist 63(3):160.

Matzler, Kurt, Birgit Renzl, Julia Muller, Stephan Herting and Todd A Mooradian. 2008. “Personalitytraits and knowledge sharing.” Journal of Economic Psychology 29(3):301–313.

McCrae, Robert R. 1987. “Creativity, divergent thinking, and openness to experience.” Journal ofpersonality and social psychology 52(6):1258.

McCrae, Robert R and Angelina R Sutin. 2009. “Openness to experience.”.

McCrae, Robert R and Oliver P John. 1998. “An introduction to the five-factor model and its applica-tions.” Personality: critical concepts in psychology 60:295.

McCroskey, James C and Virginia P Richmond. 1995. “Correlates of compulsive communication: Quan-titative and qualitative characteristics.” Communication Quarterly 43(1):39–52.

McFarland, Daniel A, Dan Jurafsky and Craig Rawlings. 2013. “Making the Connection: Social Bondingin Courtship Situations.” American journal of sociology 118(6):1596–1649.

Menon, Tanya and Jeffrey Pfeffer. 2003. “Valuing internal vs. external knowledge: Explaining thepreference for outsiders.” Management Science 49(4):497–513.

Nanda, Ramana and Jesper B Sørensen. 2010. “Workplace peers and entrepreneurship.” ManagementScience 56(7):1116–1126.

Obstfeld, David. 2005. “Social networks, the tertius iungens orientation, and involvement in innovation.”Administrative science quarterly 50(1):100–130.

Oh, Hongseok, Giuseppe Labianca and Myung-Ho Chung. 2006. “A multilevel model of group socialcapital.” Academy of management review 31(3):569–582.

Oh, Hongseok, Myung-Ho Chung and Giuseppe Labianca. 2004. “Group social capital and groupeffectiveness: The role of informal socializing ties.” Academy of management journal 47(6):860–875.

Paulus, Paul B and Huei-Chuan Yang. 2000. “Idea generation in groups: A basis for creativity inorganizations.” Organizational behavior and human decision processes 82(1):76–87.

Perry-Smith, Jill E. 2006. “Social yet creative: The role of social relationships in facilitating individualcreativity.” Academy of Management journal 49(1):85–101.

Podolny, J. M. 1993. “A Status-Based Model of Market Competition.” American Journal of Sociology98(4):829–872.

Podolny, J. M. 2001. “Networks as the pipes and prisms of the market.” American Journal of Sociology107(1):33–60.

Rammstedt, Beatrice and Oliver P John. 2007. “Measuring personality in one minute or less: A 10-itemshort version of the Big Five Inventory in English and German.” Journal of research in Personality41(1):203–212.

Reagans, R. and E. W. Zuckerman. 2001. “Networks, Diversity, and Productivity: The Social Capitalof Corporate R&D Teams.” Organization Science 12(4):502.

Reagans, Ray, Ezra Zuckerman and Bill McEvily. 2004. “How to make the team: Social networks vs.demography as criteria for designing effective teams.” Administrative Science Quarterly 49(1):101–133.

31

Page 32: Sharique Hasan Rembrand Koning Stanford GSB …...Sharique Hasan Stanford GSB Rembrand Koning Stanford GSB September 29, 2015 Abstract Social interaction is thought to a ect individual

Reily, Ken, Pam Ludford Finnerty and Loren Terveen. 2009. Two peers are better than one: aggregatingpeer reviews for computing assignments is surprisingly accurate. In Proceedings of the ACM 2009international conference on Supporting group work. ACM pp. 115–124.

Reinholt, MIA, Torben Pedersen and Nicolai J Foss. 2011. “Why a central network position isn’tenough: the role of motivation and ability for knowledge sharing in employee networks.” Academy ofManagement Journal 54(6):1277–1297.

Ries, Eric. 2011. The lean startup: How today’s entrepreneurs use continuous innovation to createradically successful businesses. Random House LLC.

Sacerdote, B. 2001. “Peer Effects with Random Assignment: Results for Dartmouth Roommates*.”Quarterly Journal of Economics 116(2):681–704.

Sacerdote, Bruce. 2014. “Experimental and quasi-experimental analysis of peer effects: two stepsforward?” Annu. Rev. Econ. 6(1):253–272.

Saxenian, AnnaLee. 1996. Regional advantage. Harvard University Press.

Shane, Scott and Daniel Cable. 2002. “Network ties, reputation, and the financing of new ventures.”Management Science 48(3):364–381.

Shue, Kelly. 2013. “Executive networks and firm policies: Evidence from the random assignment ofMBA peers.” Review of Financial Studies 26(6):1401–1442.

Singh, Jasjit and Lee Fleming. 2010. “Lone inventors as sources of breakthroughs: Myth or reality?”Management Science 56(1):41–56.

Smith, Edward Bishop, Tanya Menon and Leigh Thompson. 2012. “Status differences in the cognitiveactivation of social networks.” Organization Science 23(1):67–82.

Smith, Sandra Susan. 2005. ““Don’t put my name on it”: Social Capital Activation and Job-FindingAssistance among the Black Urban Poor1.” American Journal of Sociology 111(1):1–57.

Sosa, Manuel E. 2011. “Where do creative interactions come from? The role of tie content and socialnetworks.” Organization Science 22(1):1–21.

Sutton, Robert I and Andrew Hargadon. 1996. “Brainstorming groups in context: Effectiveness in aproduct design firm.” Administrative Science Quarterly pp. 685–718.

Tortoriello, Marco. 2015. “The social underpinnings of absorptive capacity: The moderating effectsof structural holes on innovation generation based on external knowledge.” Strategic ManagementJournal 36(4):586–597.

Tortoriello, Marco and David Krackhardt. 2010. “Activating cross-boundary knowledge: The role ofSimmelian ties in the generation of innovations.” Academy of Management Journal 53(1):167–181.

Tsai, Wenpin. 2001. “Knowledge transfer in intraorganizational networks: Effects of network positionand absorptive capacity on business unit innovation and performance.” Academy of managementjournal 44(5):996–1004.

Uzzi, Brian. 1996. “The sources and consequences of embeddedness for the economic performance oforganizations: The network effect.” American sociological review pp. 674–698.

Uzzi, Brian. 1997. “Social structure and competition in interfirm networks: The paradox of embedded-ness.” Administrative science quarterly pp. 35–67.

Woolley, Anita Williams, Christopher F Chabris, Alex Pentland, Nada Hashmi and Thomas W Malone.2010. “Evidence for a collective intelligence factor in the performance of human groups.” science330(6004):686–688.

32

Page 33: Sharique Hasan Rembrand Koning Stanford GSB …...Sharique Hasan Stanford GSB Rembrand Koning Stanford GSB September 29, 2015 Abstract Social interaction is thought to a ect individual

Wuchty, Stefan, Benjamin F Jones and Brian Uzzi. 2007. “The increasing dominance of teams inproduction of knowledge.” Science 316(5827):1036–1039.

Zahra, Shaker A and Gerard George. 2002. “Absorptive capacity: A review, reconceptualization, andextension.” Academy of management review 27(2):185–203.

Zimmerman, D.J. 2003. “Peer effects in academic outcomes: Evidence from a natural experiment.”Review of Economics and Statistics 85(1):9–23.

33

Page 34: Sharique Hasan Rembrand Koning Stanford GSB …...Sharique Hasan Stanford GSB Rembrand Koning Stanford GSB September 29, 2015 Abstract Social interaction is thought to a ect individual

1 Appendix

1.1 Individual Idea Generation

To provide context for the nature of the ideas generated during the individual idea generation process,we, present examples of raw ideas generated immediately after the three randomized interviews thatwere rated highly as well as poorly on the three dimensions of business potential, buy likelihood andnovelty.

Examples of highly rated ideas include:

Feast on demand lets the wedding planners minimise food wastage during the feasts in theevents. Through this app, the wedding planners can generate a link and forward it to allthe guests. On opening that link, the guests are confronted with a set of choices of fooditems/dishes they wish to consume during the event. After the guests give their preferences,the wedding planner gets the data and can arrange the food according to these estimates.Also, the dishes with low preference can be eliminated to the reduce wastage.

Behavioral analysis of bride and grooms online profiles on key social networks. This couldbe done exclusively by a company which would give a detailed analysis by psychologists. Thiswould definitely aid the match-making process, making it more thorough.

Renting of Wedding Dresses. Most women don’t sell off jewelry bought, but dresses cannotbe re-worn. Since branding is all that matters when it comes to second hand, the dressescould be dry washed and repacked in bags and delivered.

Examples of ideas that received low ratings include:

PERSONALISED CARDS. [my interviewee] said that it gets to be highly painful to writenames on cards and thus I propose that an agency that sends personalised cards and trackswhether they have reached.

connectivity of app event and fb event is a nice way to spread info easily

Build an app that would give users a complete guide on personal grooming tips for weddings(from deciding on what to wear to how to wear the make-up to how to carry yourself,etc)customized according to the user’s built, complexion, and personality.

34

Page 35: Sharique Hasan Rembrand Koning Stanford GSB …...Sharique Hasan Stanford GSB Rembrand Koning Stanford GSB September 29, 2015 Abstract Social interaction is thought to a ect individual

1.2 Team Submissions

To provide reference points for how evaluators rated the final team submissions, we provide examplesof submissions in the top, middle, and bottom quartiles of submissions in terms of total score.

An example of a submission in the top quartile was a prototype for mobile app called “SnappilyWed.” The team’s description of the product is:

Your guests use smart phones to take photos at the wedding but don’t share them with you.For you it’s a loss of precious memories. Our App solves the problem by allowing yourGuests to take pictures and directly saving them on the cloud. Don’t miss out on yourwedding. Capture and retain every photo taken by Everybody at your wedding (be it youruncle playing with your nieces or your brother taking photos of the food served).The marryingcouple (you or the person maintaining your account) will have access to these pictures andwill retain and share the ones which are great, while discarding the rest, for your loved onesto view.

Their splash page depicted in figure 4, is clear and visually appealing:

[Figure 4 about here.]

An example of a submission in near the 50th percentile is “Tender my Wedding.” The team describestheir idea as:

TenderMyWedding is a platform which turns the process of finding vendors for a weddingupside down. Rather than the customer looking for vendors for their wedding needs, welet Vendors look for them. All they do is simply post their requirements with budget andwithin no time, top service providers from everywhere would be competing to get them astheir customer. It’s a win-win as you get multiple cost-effective quotes for the requirementswithout stepping out of your home and Vendors get new business.

Their splash page submission, depicted in figure 5:

[Figure 5 about here.]

An example of a submission in the bottom quartile of the ratings is “Invite My Pals” which isdescribed as:

Invite My Pals makes inviting people a much easier task with superb efficiency! Be it weddingor any other occasion, using this app you can send invitations to people that will not justdirectly reach them but also would let you keep track of how many people are going to joinyou on your day. With the video invites and e-cards best suiting to your taste you sendinvitations in more personalised way than ever before!!

Their splash page submission, depicted in figure 6:

[Figure 6 about here.]

35

Page 36: Sharique Hasan Rembrand Koning Stanford GSB …...Sharique Hasan Stanford GSB Rembrand Koning Stanford GSB September 29, 2015 Abstract Social interaction is thought to a ect individual

List of Figures

1 Photo of participant entering idea into system. . . . . . . . . . . . . . . 372 Note taking sheet for each empathy interview. . . . . . . . . . . . . . . . 383 Visual summary of experimental procedure and data collection.. . . . . 394 Splash page for submission in the top quartile—Snappily Wed. . . . . . 405 Splash page for submission in the middle quartile—Tender my Wedding. 416 Splash page for submission in the bottom quartile—Invite My Pals. . . . 42

36

Page 37: Sharique Hasan Rembrand Koning Stanford GSB …...Sharique Hasan Stanford GSB Rembrand Koning Stanford GSB September 29, 2015 Abstract Social interaction is thought to a ect individual

Figure 1: Photo of participant entering idea into system.

37

Page 38: Sharique Hasan Rembrand Koning Stanford GSB …...Sharique Hasan Stanford GSB Rembrand Koning Stanford GSB September 29, 2015 Abstract Social interaction is thought to a ect individual

Figure 2: Note taking sheet for each empathy interview.

38

Page 39: Sharique Hasan Rembrand Koning Stanford GSB …...Sharique Hasan Stanford GSB Rembrand Koning Stanford GSB September 29, 2015 Abstract Social interaction is thought to a ect individual

Figure 3: Visual summary of experimental procedure and data collection..

39

Page 40: Sharique Hasan Rembrand Koning Stanford GSB …...Sharique Hasan Stanford GSB Rembrand Koning Stanford GSB September 29, 2015 Abstract Social interaction is thought to a ect individual

Figure 4: Splash page for submission in the top quartile—Snappily Wed.

40

Page 41: Sharique Hasan Rembrand Koning Stanford GSB …...Sharique Hasan Stanford GSB Rembrand Koning Stanford GSB September 29, 2015 Abstract Social interaction is thought to a ect individual

Figure 5: Splash page for submission in the middle quartile—Tender my Wedding.

41

Page 42: Sharique Hasan Rembrand Koning Stanford GSB …...Sharique Hasan Stanford GSB Rembrand Koning Stanford GSB September 29, 2015 Abstract Social interaction is thought to a ect individual

Figure 6: Splash page for submission in the bottom quartile—Invite My Pals.

42

Page 43: Sharique Hasan Rembrand Koning Stanford GSB …...Sharique Hasan Stanford GSB Rembrand Koning Stanford GSB September 29, 2015 Abstract Social interaction is thought to a ect individual

List of Tables

1 Summary statistics for ideation analysis . . . . . . . . . . . . . . . . . . 442 Balance test examining relationship between peers and pre-conversation

ideas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 453 Summary statistics for feedback analysis . . . . . . . . . . . . . . . . . . 464 Balance test examining correlation between self and peer’s traits for feed-

back analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 475 Do conversations with extroverted peers increase idea quality? . . . . . . 486 Do extroverted peers increase the quality of ideas all all dimensions? . . 497 Does a focal innovators openness moderate the effect of extroverted peers?

508 The moderating effect of focal innovator’s openness on idea quality. . . . 519 Robustness to controlling for other peer characteristics. . . . . . . . . . 5210 Robustness to controlling for whether evaluator and focal innovator have

a relationship (pre-treatment). . . . . . . . . . . . . . . . . . . . . . . . 5311 The moderating effect of team openness and internal dynamics on the

effect of extroverted peers. . . . . . . . . . . . . . . . . . . . . . . . . . . 54

43

Page 44: Sharique Hasan Rembrand Koning Stanford GSB …...Sharique Hasan Stanford GSB Rembrand Koning Stanford GSB September 29, 2015 Abstract Social interaction is thought to a ect individual

Table 1: Summary statistics for ideation analysiscount mean sd min max

Idea Quality 1150 7.548 2.824 3.000 15.000business 1203 2.476 1.055 1.000 5.000buy 1352 2.578 1.106 1.000 5.000novelty 1765 2.405 1.097 1.000 5.000Extraversion (Peer) 2071 -0.000 1.000 -3.299 3.071Extraversion (Self) 2071 -0.000 1.000 -2.955 2.399Ability (Peer) 2071 -0.000 1.000 -2.322 2.244Ability (Self) 2071 0.000 1.000 -2.387 1.622Pre-treatment Idea Quality 2052 0.000 1.000 -2.296 4.438Agreeableness (Peer) 2071 0.000 1.000 -2.347 2.602Conscientious (Peer) 2071 -0.000 1.000 -2.548 3.376Neuroticism (Peer) 2071 -0.000 1.000 -3.340 2.315Openness (Peer) 2071 -0.000 1.000 -2.054 2.529

Observations 2071

44

Page 45: Sharique Hasan Rembrand Koning Stanford GSB …...Sharique Hasan Stanford GSB Rembrand Koning Stanford GSB September 29, 2015 Abstract Social interaction is thought to a ect individual

Table 2: Balance test examining relationship between peers and pre-conversation ideas(1) (2) (3) (4) (5)

Business (pre) Buy (pre) Novelty (pre) Openness (Self) Extraversion (Self)

Extraversion (Peer) -0.076 0.003 -0.041 0.008 0.048(0.082) (0.098) (0.112) (0.078) (0.082)

Observations 107 107 107 108 108ll -146.566 -147.296 -154.197 -156.459 -154.665

Standard errors in parenthesesAll tests are two tailed. Standard errors clustered at the individual innovator level.∗ p < .1, ∗∗ p < 0.05, ∗∗∗ p < 0.01

45

Page 46: Sharique Hasan Rembrand Koning Stanford GSB …...Sharique Hasan Stanford GSB Rembrand Koning Stanford GSB September 29, 2015 Abstract Social interaction is thought to a ect individual

Table 3: Summary statistics for feedback analysiscount mean sd min max

Idea Quality 1150 7.548 2.824 3.000 15.000business 1203 2.476 1.055 1.000 5.000buy 1352 2.578 1.106 1.000 5.000novelty 1765 2.405 1.097 1.000 5.000Extraversion (Peer) 2071 -0.000 1.000 -3.299 3.071Extraversion (Self) 2071 -0.000 1.000 -2.955 2.399Ability (Peer) 2071 -0.000 1.000 -2.322 2.244Ability (Self) 2071 0.000 1.000 -2.387 1.622Pre-treatment Idea Quality 2052 0.000 1.000 -2.296 4.438Agreeableness (Peer) 2071 0.000 1.000 -2.347 2.602Conscientious (Peer) 2071 -0.000 1.000 -2.548 3.376Neuroticism (Peer) 2071 -0.000 1.000 -3.340 2.315Openness (Peer) 2071 -0.000 1.000 -2.054 2.529

Observations 2071

46

Page 47: Sharique Hasan Rembrand Koning Stanford GSB …...Sharique Hasan Stanford GSB Rembrand Koning Stanford GSB September 29, 2015 Abstract Social interaction is thought to a ect individual

Table 4: Balance test examining correlation between self and peer’s traits for feedback analysis.(1) (2)

Team’s Avg Extraversion Team’s Avg Openness

Empathy Peers’ Extraversion (Teamlevel) 0.016 0.042(0.181) (0.080)

Constant -0.010 0.015(0.162) (0.091)

Observations 40 40R2 0.000 0.005

Standard errors in parenthesesAll tests are two tailed. Standard errors clustered at the individual innovator level.∗ p < .1, ∗∗ p < 0.05, ∗∗∗ p < 0.01

47

Page 48: Sharique Hasan Rembrand Koning Stanford GSB …...Sharique Hasan Stanford GSB Rembrand Koning Stanford GSB September 29, 2015 Abstract Social interaction is thought to a ect individual

Table 5: Do conversations with extroverted peers increase idea quality?(1) (2) (3) (4) (5) (6)

Idea QualityExtraversion (Peer) 0.153∗∗ 0.170∗∗∗ 0.169∗∗∗ 0.177∗∗∗

(0.060) (0.063) (0.062) (0.060)

Extraversion (Self) -0.093 -0.091 -0.082(0.064) (0.064) (0.067)

Ability (Peer) 0.012 0.015 0.019 0.047(0.062) (0.061) (0.057) (0.056)

Ability (Self) -0.042 -0.021 -0.018(0.065) (0.062) (0.061)

Pre-treatment Idea Quality 0.122∗

(0.066)

Observations 1150 1150 1150 1150 1150 1141Innovators 107 107 107 107 107 107ll -2735.939 -2734.451 -2740.497 -2740.161 -2734.307 -2712.812

All tests are two tailed. Standard errors clustered at the individual innovator level.∗ p < .1, ∗∗ p < 0.05, ∗∗∗ p < 0.01

48

Page 49: Sharique Hasan Rembrand Koning Stanford GSB …...Sharique Hasan Stanford GSB Rembrand Koning Stanford GSB September 29, 2015 Abstract Social interaction is thought to a ect individual

Table 6: Do extroverted peers increase the quality of ideas all all dimensions?(1) (2) (3)

business buy novelty

mainExtraversion (Peer) 0.141∗∗ 0.165∗∗∗ 0.140∗∗

(0.063) (0.054) (0.061)

Extraversion (Self) -0.053 -0.060 -0.061(0.063) (0.064) (0.066)

Ability (Peer) -0.008 0.061 0.009(0.055) (0.058) (0.058)

Ability (Self) 0.007 0.005 -0.035(0.063) (0.056) (0.057)

Business (pre) 0.050(0.058)

Buy (pre) 0.106(0.066)

Novelty (pre) 0.160∗∗∗

(0.054)

Observations 1192 1342 1749Innovators 107 107 107ll -1698.504 -1965.674 -2516.945

All tests are two tailed. Standard errors clustered at the individual innovator level.∗ p < .1, ∗∗ p < 0.05, ∗∗∗ p < 0.01

49

Page 50: Sharique Hasan Rembrand Koning Stanford GSB …...Sharique Hasan Stanford GSB Rembrand Koning Stanford GSB September 29, 2015 Abstract Social interaction is thought to a ect individual

Table 7: Does a focal innovators openness moderate the effect of extroverted peers?(1) (2) (3) (4) (5) (6)

business buy novelty business buy novelty

mainExtraversion (Peer) 0.191∗∗∗ 0.224∗∗∗ 0.172∗∗∗ 0.179∗∗∗ 0.219∗∗∗ 0.200∗∗∗

(0.055) (0.049) (0.060) (0.058) (0.058) (0.062)

Extraversion (Self) -0.057 -0.084 -0.057 -0.044 -0.067 -0.048(0.060) (0.058) (0.072) (0.064) (0.064) (0.069)

Ability (Peer) 0.042 0.088 0.036 0.031 0.083 0.042(0.055) (0.056) (0.062) (0.055) (0.061) (0.063)

Ability (Self) 0.022 0.011 -0.014 0.033 0.037 0.010(0.058) (0.053) (0.057) (0.064) (0.058) (0.055)

Pre-treatment Idea Quality 0.121∗∗ 0.088∗ 0.134∗∗ 0.111∗∗ 0.076 0.124∗

(0.052) (0.052) (0.066) (0.055) (0.053) (0.064)

Openness (Self) -0.078 -0.070 -0.077(0.053) (0.055) (0.050)

Extraversion (Peer) × Openness (Self) 0.195∗∗∗ 0.232∗∗∗ 0.122∗∗

(0.072) (0.056) (0.059)

Openness (Peer) -0.092 -0.119∗∗ -0.164∗∗∗

(0.063) (0.060) (0.061)

Openness (Peer) × Extraversion (Self) -0.061 -0.102∗ -0.043(0.063) (0.060) (0.062)

Observations 1192 1342 1749 1192 1342 1749Innovators 107 107 107 107 107 107ll -1691.823 -1959.315 -2515.840 -1695.209 -1963.205 -2513.369

All tests are two tailed. Standard errors clustered at the individual innovator level.∗ p < .1, ∗∗ p < 0.05, ∗∗∗ p < 0.01

50

Page 51: Sharique Hasan Rembrand Koning Stanford GSB …...Sharique Hasan Stanford GSB Rembrand Koning Stanford GSB September 29, 2015 Abstract Social interaction is thought to a ect individual

Table 8: The moderating effect of focal innovator’s openness on idea quality.(1) (2)

Idea Quality Idea Quality

Idea QualityExtraversion (Peer) 0.281∗∗∗ 0.250∗∗∗

(0.052) (0.060)

Extraversion (Self) -0.072 -0.055(0.061) (0.064)

Ability (Peer) 0.117∗ 0.104∗

(0.061) (0.060)

Ability (Self) 0.046 0.051(0.066) (0.067)

Pre-treatment Idea Quality 0.119∗∗ 0.111∗

(0.054) (0.057)

Openness (Self) -0.126∗∗ -0.108∗

(0.057) (0.062)

Openness (Peer) -0.167∗∗ -0.180∗∗

(0.068) (0.071)

Extraversion (Peer) × Openness (Self) 0.184∗∗∗

(0.065)

Openness (Peer) × Extraversion (Self) -0.056(0.063)

Observations 1141 1141Innovators 107 107ll -2703.955 -2707.460

All tests are two tailed. Standard errors clustered at the individual innovator level.∗ p < .1, ∗∗ p < 0.05, ∗∗∗ p < 0.01

51

Page 52: Sharique Hasan Rembrand Koning Stanford GSB …...Sharique Hasan Stanford GSB Rembrand Koning Stanford GSB September 29, 2015 Abstract Social interaction is thought to a ect individual

Table 9: Robustness to controlling for other peer characteristics.

(1) (2) (3) (4) (5)Idea Quality Idea Quality Idea Quality Idea Quality Idea Quality

Idea QualityExtraversion (Peer) 0.222∗∗∗ 0.282∗∗∗ 0.260∗∗∗ 0.204∗∗∗ 0.218∗∗∗

(0.051) (0.052) (0.050) (0.052) (0.053)

Extraversion (Self) -0.092 -0.076 -0.096 -0.072 -0.078(0.061) (0.062) (0.066) (0.060) (0.059)

Ability (Peer) 0.084 0.106∗ 0.073 0.091 0.086(0.058) (0.059) (0.063) (0.060) (0.054)

Ability (Self) -0.004 0.043 -0.020 -0.017 -0.008(0.058) (0.061) (0.057) (0.054) (0.057)

Pre-treatment Idea Quality 0.129∗∗ 0.138∗∗ 0.139∗∗ 0.120∗∗ 0.129∗∗

(0.062) (0.061) (0.058) (0.061) (0.062)

Openness (Self) -0.089 -0.117∗∗ -0.084 -0.105∗ -0.065(0.058) (0.056) (0.056) (0.057) (0.058)

Extraversion (Peer) × Openness (Self) 0.189∗∗∗ 0.135∗ 0.214∗∗ 0.185∗∗∗ 0.159∗∗

(0.069) (0.073) (0.090) (0.070) (0.070)

Openness (Peer) -0.157∗∗∗

(0.061)

Openness (Peer) × Openness (Self) 0.122(0.075)

Neuroticism (Peer) 0.121∗

(0.066)

Neuroticism (Peer) × Openness (Self) 0.023(0.088)

Conscientious (Peer) -0.103∗∗

(0.050)

Conscientious (Peer) × Openness (Self) -0.112∗∗

(0.054)

Agreeableness (Peer) -0.024(0.051)

Agreeableness (Peer) × Openness (Self) 0.073∗

(0.037)

Observations 1141 1141 1141 1141 1141Innovators 107 107 107 107 107ll -2707.965 -2702.455 -2705.436 -2704.661 -2706.432

All tests are two tailed. Standard errors clustered at the individual innovator level.∗ p < .1, ∗∗ p < 0.05, ∗∗∗ p < 0.01

52

Page 53: Sharique Hasan Rembrand Koning Stanford GSB …...Sharique Hasan Stanford GSB Rembrand Koning Stanford GSB September 29, 2015 Abstract Social interaction is thought to a ect individual

Table 10: Robustness to controlling for whether evaluator and focal innovator have a relationship(pre-treatment).

(1) (2) (3) (4) (5)Idea Quality Idea Quality Idea Quality Idea Quality Idea Quality

Idea QualityExtraversion (Peer) 0.229∗∗∗ 0.228∗∗∗ 0.229∗∗∗ 0.230∗∗∗ 0.233∗∗∗

(0.054) (0.054) (0.053) (0.054) (0.055)

Extraversion (Self) -0.105 -0.109∗ -0.107 -0.106 -0.112∗

(0.065) (0.066) (0.065) (0.065) (0.064)

Ability (Peer) 0.083 0.084 0.085 0.085 0.068(0.060) (0.060) (0.061) (0.060) (0.059)

Ability (Self) -0.006 -0.006 -0.006 -0.006 -0.006(0.059) (0.059) (0.059) (0.059) (0.062)

Pre-treatment Idea Quality 0.145∗∗ 0.146∗∗ 0.146∗∗ 0.145∗∗ 0.152∗∗

(0.062) (0.062) (0.062) (0.062) (0.061)

Extraversion (Peer) × Openness (Self) 0.186∗∗∗ 0.186∗∗∗ 0.186∗∗∗ 0.186∗∗∗ 0.173∗∗∗

(0.068) (0.068) (0.068) (0.068) (0.066)

Evaluator knows innovator -0.035(0.197)

Evaluator is friends with innovator 0.402(0.500)

Innovator sought advice from evaluator 0.338(0.394)

Same ice-breaker table 0.228(0.377)

Observations 1123 1123 1123 1123 1094Innovators 107 107 107 107 107ll -2652.670 -2652.443 -2652.479 -2652.511 -2588.157

All tests are two tailed. Standard errors clustered at the individual innovator level.∗ p < .1, ∗∗ p < 0.05, ∗∗∗ p < 0.01

53

Page 54: Sharique Hasan Rembrand Koning Stanford GSB …...Sharique Hasan Stanford GSB Rembrand Koning Stanford GSB September 29, 2015 Abstract Social interaction is thought to a ect individual

Table 11: The moderating effect of team openness and internal dynamics on the effect of extrovertedpeers.

(1) (2) (3) (4) (5)Total Novelty Business Prototype Total

Empathy Peers’ Extraversion (Teamlevel) 0.107∗∗ 0.163∗∗ 0.029 0.097∗ 0.079(0.053) (0.072) (0.046) (0.056) (0.055)

Team’s Avg Ability 0.101∗ 0.072 0.087∗ 0.181∗∗∗ 0.080(0.050) (0.066) (0.047) (0.060) (0.051)

Empathy Peers’ Ability (Teamlevel) -0.030 -0.074 -0.063∗ 0.050 -0.008(0.047) (0.057) (0.033) (0.060) (0.049)

Team’s Avg Extraversion -0.119∗∗ -0.073 -0.183∗∗∗ -0.022 -0.116∗∗

(0.051) (0.068) (0.043) (0.064) (0.051)

Team’s Avg Openness 0.107 0.073 0.040 0.097 0.114(0.087) (0.118) (0.073) (0.112) (0.081)

(Teamlevel) Extraversion (Peer) * Openness (Team) 0.302∗∗∗ 0.288∗ 0.352∗∗∗ 0.103 0.298∗∗∗

(0.093) (0.147) (0.072) (0.089) (0.090)

Member Cohesiveness (Team) -0.048(0.033)

Member Cohesiveness (Team) x Open (Team) x Extrav (Peer) 0.488∗∗

(0.228)

Constant 2.988∗∗∗ 2.989∗∗∗ 2.986∗∗∗ 3.000∗∗∗ 2.957∗∗∗

(0.045) (0.058) (0.046) (0.063) (0.048)

Observations 577 577 577 577 577Teams 40 40 40 40 40R2 0.044 0.032 0.039 0.032 0.053

All tests are two tailed. Standard errors clustered at the team level.∗ p < .1, ∗∗ p < 0.05, ∗∗∗ p < 0.01

54