ArticlePDF Available

A Social Network Perspective on the Bamboo Ceiling: Ethnic Homophily Explains Why East Asians but Not South Asians Are Underrepresented in Leadership in Multiethnic Environments

November 2021
Journal of Personality and Social Psychology 122(6):959–982

Authors:

Massachusetts Institute of Technology

In the United States, Asians appear disproportionately underrepresented in leadership roles, a puzzling phenomenon known as the “Bamboo Ceiling” (Hyun, 2005; Lu et al., 2020). We advance a social network explanation for this phenomenon: ethnic homophily. We theorize that East Asians (e.g., ethnic Chinese)—but not South Asians (e.g., ethnic Indians)—are less likely than other ethnicities to emerge as leaders in multiethnic environments partly because East Asians socialize more with ethnic ingroup members (other East Asians). Analyzing a survey of 54,620 Juris Doctor (JD) students from 124 U.S. law schools, Study 1 revealed that East Asians had the highest ethnic homophily of all ethnicities. Studies 2 and 3 examined friendship networks and leadership emergence in 11 class sections of new Master of Business Administration (MBA) students in a U.S. business school, and found that East Asians were the least likely to be nominated and elected as leaders. Social network analysis revealed that, compared to South Asians, Latinos, and Whites, East Asians exhibited higher ethnic homophily, which mediated their lower leadership emergence. These effects occurred for both East Asian internationals and East Asian Americans, and were robust after accounting for variables such as assertiveness (parallel mediator), network centrality, English proficiency, demographics, and personality. By integrating social network analysis into psychology, we identify ethnic homophily as one reason why the Bamboo Ceiling exists for East Asians but not South Asians. Moreover, by uncovering the negative link between ethnic homophily and leadership emergence, our research suggests that bonding with people from different ethnic backgrounds can facilitate individuals’ leadership emergence in multiethnic environments.

Study 3: Simultaneous Mediation Analysis for Leader Nominations

…

Study 2: Descriptive Statistics and Pearson Correlations

…

Figures - uploaded by Jackson G. Lu

Content may be subject to copyright.

Content uploaded by Jackson G. Lu

Content may be subject to copyright.

INNOVATIONS IN SOCIAL PSYCHOLOGY

A Social Network Perspective on the Bamboo Ceiling:

Ethnic Homophily Explains Why East Asians but Not South Asians

Are Underrepresented in Leadership in Multiethnic Environments

Jackson G. Lu

Sloan School of Management, Massachusetts Institute of Technology

In the United States, Asians appear disproportionately underrepresented in leadership roles, a puzzling

phenomenon known as the “Bamboo Ceiling”(Hyun, 2005;Lu et al., 2020). We advance a social network

explanation for this phenomenon: ethnic homophily. We theorize that East Asians (e.g., ethnic Chinese)—but

not South Asians (e.g., ethnic Indians)—are less likely than other ethnicities to emerge as leaders in

multiethnic environments partly because East Asians socialize more with ethnic ingroup members (other

East Asians). Analyzing a survey of 54,620 Juris Doctor (JD) students from 124 U.S. law schools, Study 1

revealed that East Asians had the highest ethnic homophily of all ethnicities. Studies 2 and 3 examined

friendship networks and leadership emergence in 11 class sections of new Master of Business Administration

(MBA) students in a U.S. businessschool, and found that EastAsians were the least likely tobe nominated and

elected as leaders. Social network analysis revealed that, compared to South Asians, Latinos, and Whites, East

Asians exhibited higher ethnic homophily, which mediated their lower leadership emergence. These effects

occurred for both East Asian internationals and East Asian Americans, and were robust after accounting for

variables suchas assertiveness (parallel mediator), network centrality, English proﬁciency, demographics, and

personality. By integrating social network analysis into psychology, we identify ethnic homophily as one

reason why the Bamboo Ceiling exists for East Asians but not South Asians. Moreover, by uncovering the

negative link between ethnic homophily and leadership emergence, our research suggests that bonding with

people from different ethnic backgrounds can facilitate individuals’leadership emergence in multiethnic

environments.

Keywords: culture, Bamboo Ceiling, leadership, social network, homophily

Well-educated and prosperous, Asians are called the “model

minority”in the United States. On average, Asians have the highest

educational attainment (Hsin & Xie, 2014;Lee & Zhou, 2015;

Ryan & Bauman, 2016), lowest unemployment rate (United States

Department of Labor, 2019), and highest median income of all

ethnic groups (Sakamoto et al., 2009;U.S. Census Bureau, 2018).

However, Asians appear disproportionately underrepresented in

leadership roles in the United States (Chin, 2020;Gee & Peck,

2018;Sy et al., 2017). This puzzling phenomenon has been termed

the “Bamboo Ceiling”(Hyun, 2005;Lu et al., 2020), in reference

to the fact that bamboo plays “an important economic and cultural

role across Asia, with the world’s largest bamboo areas in South

Asia and East Asia”(Lu et al., 2020, p. 4590).

To better understand this Bamboo Ceiling phenomenon, we move

beyond existing research to offer a social network explanation:

ethnic homophily (McPherson et al., 2001). We propose that East

Asians (EAs; e.g., ethnic Chinese, Japanese)—but not South Asians

(SAs; e.g., ethnic Indians, Pakistanis)—are less likely than other

ethnicities to emerge as leaders in multiethnic environments partly

because EAs socialize more with ethnic ingroup members (i.e.,

other EAs).

Our proposition is supported by a survey study of

This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Jackson G. Lu https://orcid.org/0000-0002-0144-9171

I thank MIT’s B-Lab, Shih-Chun Chien, Chad Christensen, Roberto

Fernandez, Adam Galinsky, Peter Jin, Shinobu Kitayama, Tucker Kuman,

Jung Won Lee, Goodwin Liu, Priya Mehla, Michael Morris, Jacquelyn

Petzold, Ray Reagans, Basima Tewﬁk, Changlan Wang, David Min-heng

Wang, Nate Wilmers, Xian Yang, Jingjing Yao, JoAnne Yates, Siyu Yu,

and Ezra Zuckerman for their helpful feedback or assistance. An earlier

version of this article was selected as a ﬁnalist for the Phillips and Nadkarni

Award for Outstanding Paper on Diversity and Cognition (Academy of

Management) and nominated for the Carolyn Dexter Award for the Best

International Paper (Academy of Management). Data from the Law School

Survey of Student Engagement (LSSSE) were used with permission from

the Indiana University Center for Postsecondary Research (CPR). The

opinions and conclusions contained in this article are those of the author

and do not necessarily reﬂect the position or policy of LSSSE, participat-

ing schools, or CPR.

Correspondence concerning this article should be addressed to Jackson G.

Lu, Sloan School of Management, Massachusetts Institute of Technology, 100

Main St, Cambridge, MA 02142, United States. Email: lu18@mit.edu

Because other Asian subgroups, including Southeast Asians (e.g., ethnic

Indonesians, Singaporeans, Vietnamese) and Central Asians (e.g., ethnic

Kazakhs, Uzbeks), have much smaller populations in the United States, our

studies did not have sufﬁcient samples to analyze these groups.

Journal of Personality and Social Psychology:

Attitudes and Social Cognition

ISSN: 0022-3514 https://doi.org/10.1037/pspa0000292

Template Version: 11 October 2021 ▪6:01 pm IST PSP-A-2020-1216_format_ﬁnal ▪8 November 2021 ▪3:53 pm IST

54,620 Juris Doctor (JD) students from 124 U.S. law schools and

two social network studies of Master of Business Administration

(MBA) students in a U.S. business school.

The theory section is structured as follows. First, we review the

existing literature on the Bamboo Ceiling phenomenon. Next, we

review the concept of ethnic homophily and theorize why it may

hinder individuals’leadership emergence in multiethnic environ-

ments. Finally, we offer a cultural psychological perspective on why

ethnic EAs tend to be high on ethnic homophily and thus less likely

to emerge as leaders in multiethnic environments.

The Bamboo Ceiling

In the United States, Asians appear disproportionately underrep-

resented in leadership roles across different industries (Gee & Peck,

2018;Hyun, 2005;Lu et al., 2020;Thatchenkery & Sugiyama,

2011). On Wall Street, although Asians are well represented at the

employee level, they are underrepresented at the executive level. For

example, according to a 2017 report, whereas 20.6% of employees

at JPMorgan Chase were Asian, only about 6.8% of its executives

were Asian (Bloomberg, 2017). Similarly, in U.S. law ﬁrms, only

3% of partners were Asian despite 11% of associates being Asian

(Chung et al., 2017). Even in the technology industry, where 30% of

the workforce is Asian, less than 15% of senior leaders are Asian

(Gee & Peck, 2018;The Economist, 2015).

In light of these striking descriptive statistics, researchers have

begun to examine the scope and mechanisms of this Bamboo

Ceiling phenomenon (Lu et al., 2020;Sy et al., 2010;Yu, 2020).

In terms of scope, most organizational and government statistics

lump all Asian subgroups together despite their cultural differences.

Recently, Lu et al. (2020) observed that in contrast to the paucity of

EA CEOs in the United States, plenty of SA CEOs have led well-

known companies such as Citigroup, Google, MasterCard, Micro-

soft, and PepsiCo. Across nine studies using mixed methods, these

researchers consistently found that ethnic EAs—but not ethnic

SAs—are underrepresented in leadership roles.

In terms of mechanisms behind the Bamboo Ceiling phenome-

non, Lu et al. (2020) tested three potential mechanisms—

motivation, prejudice, and assertiveness—while controlling for a

broad set of personality and demographic variables (e.g., birth

country, English proﬁciency, education, socioeconomic status

[SES]). Analyses found that EAs were as motivated as SAs and

Whites to become leaders, and EAs actually faced less prejudice

than SAs. However, EAs were lower than SAs and Whites in

assertiveness, which partly mediated EAs’disadvantage in leader-

ship emergence (Lu et al., 2020). The assertiveness mechanism is

consistent with leadership categorization theory (Lord & Maher,

1991;Shondrick et al., 2010), which posits that individuals are less

likely to emerge as leaders if their characteristics are incongruent

with the leader prototype in a given culture. EAs’low assertiveness

is incongruent with the American leadership prototype, as actively

asserting oneself signals conﬁdence and conviction in mainstream

American culture (Lu et al., 2020).

Besides assertiveness, the stereotypes of EAs as competent but

cold and foreign may also contribute to the Bamboo Ceiling

(Berdahl & Min, 2012;Cheryan & Monin, 2005;Lin et al.,

2005;Zou & Cheryan, 2017). Such stereotypes can result in

others’envy and hostility, impeding EAs’leadership emergence

(Johnson & Sy, 2016). In addition, the underrepresentation of EA

leaders can self-perpetuate: Because there are numerically few EA

leaders in the United States, EAs are typically not associated with

the image of “leader”(Rosette et al., 2008;Sy et al., 2010). As a

result, EAs are less likely to be selected as leaders, which in turn

reinforces their “non-leader”image; this can produce a vicious cycle

that exacerbates the underrepresentation of EA leaders.

While the above perspectives have provided insights into the Bamboo

Ceiling, our understanding of this phenomenon is far from complete.

In particular, while these perspectives have examined characteristics

(e.g., assertiveness) and stereotypes of Asian individuals, they have not

yet accounted for social relational dynamics—even though leadership

emergence is inherently a social and relational process (Balkundi &

Kilduff, 2005;Carter et al., 2015). To ﬁll this knowledge gap, we

break new ground by advancing a social network explanation for

the Bamboo Ceiling phenomenon: ethnic homophily.

Ethnic Homophily

Homophily—the sociological principle that “birds of a feather

ﬂock together”—has been studied across a wide range of settings,

attributes, and relationships (Fernandez et al., 2000;Gibbons &

Olk, 2003;Ibarra, 1992,1993;Lawrence & Shah, 2020;McPherson

et al., 2001;Melamed et al., 2020;Mollica et al., 2003;Reagans,

2005). More speciﬁcally, ethnic homophily is deﬁned as the

preference for interacting with individuals of the same ethnicity

(Leszczensky & Pink, 2019;Smith et al., 2016). In their inﬂuential

review, McPherson et al. (2001) noted that ethnic homophily tends

to be more salient than other types of homophily (e.g., age homo-

phily, gender homophily) in social environments.

Past research suggests that ethnic homophily is a double-edged

sword. On the one hand, same-ethnicity relationships can provide

social support for ethnic minorities (Friedman et al., 1998;James,

2000), because sharing the same ethnic background “increases ease

of communication, improves predictability of behavior, and fosters

relationships of trust and reciprocity”(Ibarra, 1993, p. 61). For

example, Friedman et al. (1998) found that Black network groups

had a positive effect on Black managers’career optimism. More-

over, ethnic homophily can accelerate the diffusion of ideas and

practices among ethnic ingroups because of higher trust and ease of

communication (Rogers, 2003). On the other hand, ethnic homo-

phily also has downsides. First, it restrains ethnic minorities’access

to diverse information and resources outside their ethnic networks

(Ertug et al., 2018;Mollica et al., 2003;Xie & Gough, 2011).

Second, ethnic homophily can result in the marginalization of ethnic

minorities in a social environment (Mehra et al., 1998), which

exacerbates ethnic inequality (Lynam & Cowley, 2007). Third,

ethnic homophily precludes the potential beneﬁts of interethnic

friendships, such as increased creativity (Lu et al., 2017;Maddux

et al., 2021), reduced prejudice (Pettigrew & Tropp, 2006), and

enhanced well-being (Mendoza-Denton & Page-Gould, 2008).

Despite these ﬁndings, little research has examined the effect of

ethnic homophily on leadership emergence in multiethnic environ-

ments. This is a notable knowledge gap because societies and

organizations are increasingly multiethnic (Pew Research Center,

2018); it is thus important to understand how individuals can emerge

as leaders in multiethnic environments. To ﬁll this knowledge gap,

we next theorize why ethnic homophily can hinder individuals’

leadership emergence in multiethnic environments.

This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

2LU

Template Version: 11 October 2021 ▪6:01 pm IST PSP-A-2020-1216_format_ﬁnal ▪8 November 2021 ▪3:53 pm IST

Ethnic Homophily Hinders Leadership

Emergence in Multiethnic Environments

Leadership emergence is inherently a social and relational

process (Judge et al., 2002). People are more likely to perceive

and select a person as their leader if they believe that the person has

suitable characteristics (Edwards, 2008) and can represent their

values and interests (Platow & Van Knippenberg, 2001;Reicher

et al., 2018). Drawing on two theoretical perspectives that high-

light such relational processes, we propose that ethnic homophily

can hinder individuals’leadership emergence in multiethnic

environments.

First, the person–environment ﬁt perspective (Edwards, 2008;

Kristof-Brown et al., 2005;Lu et al., in press) suggests that

individuals are more likely to emerge as leaders when there is a

ﬁt between personal characteristics and environmental character-

istics. A multiethnic environment calls for leaders who can bond

with members of different ethnic groups. Compared to ethnically

homophilous leaders, ethnically heterophilous leaders are a better

ﬁt for multiethnic environments because they can serve as brokers

that bridge different ethnic groups. For example, in a multiethnic

MBA class, if a person only socializes with ethnic ingroup class-

mates, it signals low openness to ethnic diversity and a lack of

interest in interacting with ethnic outgroup members (Lisak & Erez,

2015). Conversely, if a person socializes with classmates of differ-

ent ethnicities, it signals his/her potential as an inclusive leader who

can bring together people from diverse ethnic backgrounds. This is

a reason why political candidates strive to connect with different

ethnic groups in their campaigns. As a revealing example, multiple

candidates in the 2020 U.S. presidential debates deliberately spoke

in Spanish to connect with Latino voters (TIME, 2019). Similarly,

many non-Chinese political candidates in San Francisco adopted

Chinese names to appeal to the large Chinese population there

(Green, 2016).

Second, the uncertainty reduction hypothesis posits that people

have an epistemic need to reduce uncertainty (Hogg, 2000). If I am

part of an ethnic outgroup (vs. ingroup) in the eyes of others, they

may feel uncertain whether I share their values and interests. As a

result, they may hesitate to trust me as a leader. However, if I am

ethnically heterophilous and socialize with their ethnic ingroup

member(s), then they may feel more certain that I share their values

and interests, even though we are not direct friends. As a result,

they are more likely to endorse me as a leader. Indeed, research

shows that as an individual has more indirect contact with outgroup

members via ingroup members, he/she is more likely to have future

direct contact with outgroup members because of reduced inter-

group anxiety (Wölfer et al., 2019). Relatedly, the principle of

transitivity means that if an ethnic outgroup member is not yet

my friend but is already friends with my ethnic ingroup member(s),

I may presume that this person is likely to become my friend in

the future (Flynn et al., 2010) and thus trust this person as a

leader (Hogg, 2001). This view is echoed in Simmel’s theory of

triadic closure (Simmel, 1908)andHeider’s balance theory

(Heider, 1946).

In light of these theoretical perspectives, we hypothesize that

more ethnically homophilous individuals are less likely to emerge as

leaders in multiethnic environments. That is, we hypothesize that an

individual’s ethnic homophily negatively predicts his/her leadership

emergence in multiethnic environments (Hypothesis 1).

Why Ethnic East Asians Are High on

Ethnic Homophily

Having theoretically established the negative link between ethnic

homophily and leadership emergence in multiethnic environments,

we next theorize why ethnic EAs are high on ethnic homophily and

thus less likely to emerge as leaders in multiethnic environments.

Although several studies have examined ethnic differences in

friendship homophily (Ibarra, 1995;Mehra et al., 1998;Mollica

et al., 2003;Quillian & Campbell, 2003;Zeng & Xie, 2008), due to

limited sample sizes most of them only contrasted Whites versus

non-Whites, let alone comparing among Asian subgroups. For

example, Mollica et al. (2003) examined the friendship networks

of three sections of ﬁrst-year MBA students (about 20 participants in

each section) and found that ethnic minorities exhibited higher

homophily than Whites. However, because of their small sample

of ethnic minority participants, the researchers could not compare

ethnic homophily among the different ethnic minorities (Mollica

et al., 2003). Similarly, Quillian and Campbell (2003) analyzed the

U.S. National Longitudinal Study of Adolescent Health to examine

ethnic differences in homophily, but this survey lumped all Asian

subgroups together. This is a notable limitation because, as dis-

cussed below, EA cultures differ from other cultures (including SA

cultures) in ways related to ethnic homophily.

We draw on cultural psychology research to hypothesize why

EAs tend to be more ethnically homophilous than other ethnic

groups. Speciﬁcally, we discuss cultural differences in (a) commu-

nication style, (b) ethnic heterogeneity, and (c) relational mobility.

Because our social network studies only had sufﬁcient sample sizes

for EAs, SAs, Latinos, and Whites, our theorization focuses on

cultural differences among these four groups.

Cultural Differences in Communication Style

A wealth of research suggests that EA cultures differ from

mainstream American culture on many dimensions, including

self-enhancement (Heine & Hamamura, 2007;Kitayama et al.,

1997), construction of agency (Markus et al., 2006), emotions

(Kitayama et al., 2006), and humor (Lu et al., 2019). Of particular

relevance to leadership are cultural differences in communication

style (Adair et al., 2016;Gudykunst et al., 1996). Whereas main-

stream American culture encourages assertiveness and emotional

expressivity, EA cultures emphasize humility and harmony (Lu

et al., 2020;Markus & Kitayama, 1991;Tsai et al., 2006). EA

cultures are deeply inﬂuenced by Confucianism, whereas both SA

cultures and Latino cultures have been inﬂuenced by Western

colonization (Jeffrey, 2018;Tatum, 2013). EAs tend to communi-

cate indirectly, implicitly, and nonassertively, whereas SAs tend to

communicate directly, explicitly, and assertively (Lu et al., 2020;

Nishimura et al., 2008;Sanchez-Burks et al., 2003). For example,

Nishimura et al. (2008) found that Japanese tend to be quiet, use

little body language, and think in silence, whereas Indians tend to be

lively, use overt body language, and think aloud. Compared to EA

cultures, Latino cultures emphasize “expressive displays of personal

charm, graciousness, and hospitality”in interpersonal communica-

tion (Sanchez-Burks et al., 2000, p. 175). For example, Graham

(1985) found that Brazilian negotiators said “no”more frequently

and had fewer silent periods than Japanese negotiators.

This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

A SOCIAL NETWORK PERSPECTIVE ON THE BAMBOO CEILING 3

Template Version: 11 October 2021 ▪6:01 pm IST PSP-A-2020-1216_format_ﬁnal ▪8 November 2021 ▪3:53 pm IST

Notably, Lu et al. (2020) found that both EA internationals and

EA Americans were less assertive than their SA and White counter-

parts. This ﬁnding suggests that EAs’low assertiveness is not just an

issue tied to English proﬁciency but rather has deep cultural roots:

Although EA Americans are native English speakers, their family

upbringings (e.g., dinner conversations with their EA parents) can

still imprint the Confucian values of humility and harmony

(Hofstede & Bond, 1988;Kitayama & Cohen, 2010;Markus &

Kitayama, 1991;Menon et al., 2010).

Due to such cultural differences in communication style, ethnic

EAs may be less at ease than other ethnicities with the social scene in

mainstream American culture, and thus socialize more within their

cultural comfort zone (i.e., with other EAs).

Cultural Differences in Ethnic Heterogeneity

Second, EA cultures’low ethnic heterogeneity may also contribute

to EAs’high ethnic homophily. Individuals from ethnically homo-

geneous cultures may have fewer scripts about interethnic interactions

(Smith et al., 2020). Their cognitive networks may be less ethnically

diverse, and as a result, they prefer socializing with ethnic ingroup

members in social environments. Moreover, research shows that

individuals from ethnically heterogeneous cultures mentally represent

ethnic groups as more similar to each other, because these individuals

are more likely to “realize latent and deep commonalities across

groups”(Bai et al., 2020, p. 12747). By contrast, individuals from

ethnically homogeneous cultures perceive ethnic groups as more

different from each other (Bai et al., 2020). Such perceived differ-

ences can create psychological barriers for socializing with ethnic

outgroup members (Turner & Cameron, 2016).

Unlike multiethnic cultures like the United States, EA cultures are

largely monoethnic. According to Fearon’s (2003) widely-used

Ethnic Diversity Index (0 =least diverse, 1 =most diverse), the

United States has a score of 0.491, whereas South Korea is 0.004,

Japan is 0.012, and China is 0.154. Indeed, EA countries are ranked

among the least ethnically diverse countries in the world (Fearon,

2003). Ethnic minorities comprise only about 1% of the population

in both Japan and South Korea (The World Factbook: Japan, 2021;

The World Factbook: South Korea, 2021). Similarly, as many as

91.6% of the Chinese population are Han Chinese (The World

Factbook: China, 2021). In contrast to EA cultures, SA cultures are

highly multiethnic (Ethnic Diversity Index: India =0.811, Pakistan =

0.532, Sri Lanka =0.428; Fearon, 2003). For example, India is

home to numerous ethnic and religious groups, including Hindus,

Muslims, Christians, Sikhs, Buddhists, Jains, Jews, Parsees, and

Baha’is (Sen, 2005). As Frey and Roysircar (2006) summarized,

“because diverse languages, religions, and cultural groups exist

within most South Asian countries, South Asians have learned to

accept the presence of cultural differences”(p. 218). Similarly, Roy

(2018) noted that “[South Asians’] natural experience of managing

diversity from their own very diverse ancestral country helps them to

quickly adapt in different circumstances and ride on the ladder of

success through wider acceptability from other communities”

(p. 203). Like SA cultures, Latino cultures are highly multiethnic

(Alesina et al., 2003). Latino cultures are characterized by a history

of mestizaje—the ethnic and cultural mixing of Latin American

indigenous peoples, Europeans, and Africans (Miller, 2009). For

example, the Ethnic Diversity Index scores for Brazil and Mexico,

the two most represented Latino cultures in our studies, are as high

as 0.549 and 0.542, respectively (Fearon, 2003).

Due to EA cultures’ethnic homogeneity, EAs may be less

accustomed to interacting with people from different ethnic back-

grounds and thus socialize more with other EAs in multiethnic

environments.

Cultural Differences in Relational Mobility

Third, EA cultures’low relational mobility may also contribute to

EAs’high ethnic homophily. Relational mobility refers to “how

much freedom and opportunity a society affords individuals to select

and replace interpersonal relationships based on their personal

preferences”(Yuki & Schug, 2020, p. 129). In cultures with high

relational mobility, a person’s relationships tend to be guided by

personal choices and preferences rather than concern for the ethnic

group to which one belongs (De et al., 2015;Yuki & Schug, 2012).

Indeed, research suggests that ethnocentrism increases as mobility

decreases, because ethnocentrism is adaptive in low mobility socie-

ties, which favor group-entitative behaviors (De et al., 2015). In a

recent cross-cultural study, Thomson et al. (2018) found that rela-

tional mobility (range: −0.414 to 0.414) was high in the U.S. (0.182)

and Latino cultures (e.g., Brazil =0.203, Mexico =0.359, Puerto

Rico =0.308, Venezuela =0.226), but low in EA cultures

(e.g., Hong Kong =−0.338, Japan =−0.414, Taiwan =−0.294).

Importantly, cultural differences in relational mobility can exist

within the same country (Falk et al., 2009;Zhang & Li, 2014). For

example, Falk et al. (2009) found that European Canadians reported

higher relational mobility than not only the Japanese in Japan but

also EA Canadians.

Whereas mainstream American culture encourages individuals to

socialize with whoever they “click with,”EAs may be more apt to

befriend another EA just because that person is an ethnic ingroup

member (Kim & Markus, 1999;Oishi et al., 2015). EAs who

primarily socialize with ethnic outgroups are sometimes frowned

upon by other EAs. For example, pejorative terms like “banana”and

“Twinkie”(EA on the outside, White on the inside) are used to

disparage EAs who primarily socialize with Whites (Trieu, 2019).

Indeed, EA cultures’prioritization of ingroup relationships is

sometimes accomplished at the expense of outgroup exclusion

(Yum, 1988). In the United States, many EA entrepreneurs and

professionals reside in ethnic enclaves in urban areas (Paik et al.,

2017). In contrast to the many Chinatowns, Japantowns, and Korea-

towns, the Indian population in the United States “by and large does

not live isolated in ‘Indiatowns’” (Oldenburg, 1988, p. 227).

In light of the aforementioned cultural differences in communi-

cation style, ethnic heterogeneity, and relational mobility, EAs may

prefer to socialize with ethnic ingroup members in multiethnic

environments. That is, we hypothesize that EAs tend to exhibit

high ethnic homophily in multiethnic environments (Hypothesis 2).

Hypotheses and Overview of Studies

Taken together, if EAs are high on ethnic homophily (Hypothesis 2)

and ethnic homophily negatively predicts leadership emergence in

multiethnic environments (Hypothesis 1), then EAs may be less

This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Thomson et al. (2018) did not present data on the relational mobility of

SA cultures.

4LU

Template Version: 11 October 2021 ▪6:01 pm IST PSP-A-2020-1216_format_ﬁnal ▪8 November 2021 ▪3:53 pm IST

likely to emerge as leaders in multiethnic environments. In other

words, we hypothesize that EAs are less likely to emerge as leaders

in multiethnic environments (Hypothesis 3) partly because of their

high ethnic homophily (Hypothesis 4).

To test these hypotheses, we conducted three studies. In

Study 1, we analyzed a large survey of 54,620 JD students

from 124 U.S. law schools to test whether EAs had higher ethnic

homophily than non-EAs (Hypothesis 2). Next, Studies 2 and 3

tested our entire theoretical perspective (Hypotheses 1–4) via

social network analysis. Speciﬁcally, we examined friendship

networks and leadership emergence in 11 class sections of new

MBA students in a U.S. business school. Given the recent ﬁnding

that assertiveness is an important mechanism for EAs’lower

leadership emergence (Lu et al., 2020), both Studies 2 and 3

accounted for and tested assertiveness as a parallel mediator. That

is, we tested whether ethnic homophily would explain EAs’low

leadership emergence above and beyond assertiveness. Our con-

ceptual model is displayed in Figure 1.

The present research offers signiﬁcant theoretical contributions

and practical implications. First, we extend the burgeoning literature

on the Bamboo Ceiling, a puzzling phenomenon that remains

understudied. Our research contributes to this literature by providing

insights into both the scope and mechanisms of the Bamboo Ceiling.

In terms of scope, we demonstrate that ethnic EAs—but not ethnic

SAs—are less likely than other ethnicities to emerge as leaders in

multiethnic environments. This ﬁnding contributes to cultural psy-

chology by highlighting the importance of understanding cultural

differences within the Asian ethnic umbrella. In terms of mechan-

isms, whereas past research has focused on EAs’low assertiveness

as a mechanism for their low leadership attainment (Lu et al., 2020),

we identify ethnic homophily as a social network mechanism.

Moreover, we shed light on an important question: How can

individuals emerge as leaders in multiethnic environments? By

uncovering the negative link between ethnic homophily and lead-

ership emergence, our studies suggest that bonding with people

from different ethnic backgrounds is conducive to leadership emer-

gence in multiethnic environments.

Study 1

As an initial test of our theoretical perspective, Study 1 analyzed a

large survey of JD students (N=54,620) from 124 U.S. law

schools. We tested whether EAs are higher on ethnic homophily

than non-EAs (Hypothesis 2).

Method

We obtained privileged data access to the Law School Survey of

Student Engagement (LSSSE), one of the largest surveys for law

school students. Each year, LSSSE collaborates with U.S. law

schools to invite their students to participate in the survey. Partici-

pation is voluntary and anonymous. We analyzed 4 consecutive

years of LSSSE data, which involved 124 U.S. law schools in total.

Sample

To ensure data quality, we applied two exclusion criteria, which

excluded about 2% of participants before data analysis. First, we

excluded students who self-reported being younger than 21 years

old (0.2%), because they were likely to be unrepresentative of the JD

student population or might have misreported age. Second, because

JD programs in U.S. law schools typically last 3 years, we excluded

students who self-reported being in their fourth year of JD (1.8%).

All results were robust without excluding this 2% of participants.

These exclusion criteria yielded a data set of 54,620 JD students

(56% female; M

age

=27.50, SD =6.29). Thirty-nine percent of the

students were in their ﬁrst year of law school, 32% second year, and

29% third year. One methodological strength of the LSSSE is its

distinction between EAs and SAs. Of these students, 1,514 were EA,

895 SA, 40,388 White, 5,331 Black, 5,277 Latino, 467 Middle

Eastern/Arab, 244 Native American, and the rest belonged to other

categories. Thanks to its large sample size, the study enabled us to

compare the ethnic homophily of these seven ethnic groups in the

United States.

Ethnic Homophily

The LSSSE survey measured ethnic heterophily with an item

from the Interactional Diversity Scale (Hu & Kuh, 2003;Loes et al.,

2012;Pascarella et al., 2012). Speciﬁcally, each participant was

asked how often he/she had serious conversations with students of a

different ethnicity than his/her own (1 =never, 2 =sometimes,

3=often, 4 =very often). We reverse coded the measure, such that

higher scores indicate higher ethnic homophily. While this measure

of ethnic heterophily has face validity, it is a single-item measure,

so we were cautious in interpreting results.

Control Variables

To account for academic aptitude, we controlled for each student’s

score on the Law School Admission Test (LSAT), which is a

prerequisite for U.S. law schools (LSAT score range: 120–180).

Based on law school records, LSSSE provided LSAT scores in the

following ordinal scale: 1 =120–145; 2 =146–150; 3 =151–155;

4=156–160; 5 =161–165; 6 =166–170; 7 =171–180.

Second, we controlled for whether a student was international or

American. Notably, 96% of the students were American, so English

proﬁciency was unlikely to be a confounding variable for any

observed differences in ethnic homophily. Indeed, all results

were robust when we limited the analyses to Americans only.

Moreover, we controlled for student age, gender, and class year

(ﬁrst, second, or third year). Finally, we controlled for school-

level characteristics: (a) whether a law school was public (N=53)

or private (N=71) and (b) school enrollment size (small =fewer

This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Figure 1

Conceptual Model

Note. Ethnic homophily =focal mediator; Assertiveness =parallel mediator.

A SOCIAL NETWORK PERSPECTIVE ON THE BAMBOO CEILING 5

Template Version: 11 October 2021 ▪6:01 pm IST PSP-A-2020-1216_format_ﬁnal ▪8 November 2021 ▪3:53 pm IST

than 500 students, medium =500–900 students, large =more

than 900 students).

Results

Descriptive statistics and Pearson correlations are displayed in

Table 1. Because students were nested within 124 U.S. law schools,

we conducted multilevel analyses to account for within-school

statistical dependence.

Consistent with Hypothesis 2, EAs were the most ethnically homo-

philous of the seven ethnicities. As detailed in Table 2 Model 1, EAs

(M=2.36) were signiﬁcantly more homophilous than SAs

(M=1.78, B=.57, SE =.04, p<.001), Blacks (M=2.06, B=

.28, SE =.03, p<.001), Latinos (M=2.00, B=.33, SE =.03,

p<.001), Middle Easterners (M=1.78, B=.54, SE =.05,

p<.001), Native Americans (M=2.00, B=.42, SE =.07,

p<.001), and Whites (M=2.17, B=.24, SE =.03, p<.001).

All results were robust after we accounted for the control variables

(all ps<.001; Table 2 Models 2 and 3).

Robustness Checks

Thanks to the large sample size, we repeated the analyses (with

controls) on different subsets of the data: EAs had signiﬁcantly

higher ethnic homophily than non-EAs when we examined (a) only

ﬁrst-year students, only second-year students, or only third-year

students; (b) only the 53 public schools or only the 71 private

schools; (c) only the 48 small-sized schools, only the 54 medium-

sized schools, or only the 22 large-sized schools (all ps<.05).

This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Table 1

Study 1: Descriptive Statistics and Pearson Correlations

Variable MSD12345678

1. East Asian (vs. South Asian) 0.63 0.48 —

2. Ethnic homophily 2.14 0.94 .28 —

3. LSAT score (1 =120–145, 7 =171–180) 3.23 1.40 .16 .01 —

4. Age 27.50 6.29 .12 −.01 −.12 —

5. Gender (1 =male, 0 =female) 0.44 0.50 .06 .00 .10 .07 —

6. International (1 =yes, 0 =no) 0.04 0.20 .29 .03 .00 .03 −.02 —

7. First-year student (1 =yes, 0 =no) 0.39 0.49 −.02 −.01 .01 −.10 −.02 −.00 —

8. School type (1 =public, 0 =private) 0.39 0.49 .06 .01 .23 −.03 .03 −.02 −.01 —

9. Small-sized school (1 =yes, 0 =no) 0.43 0.50 −.10 .05 −.13 .02 .03 −.05 .00 .08

Note. LSAT =Law School Admission Test.

Table 2

Study 1: Multilevel Linear Regressions Predicting Ethnic Homophily

Variable

Model 1 Model 2 Model 3

BSEBSEBSE

East Asian (reference category)

South Asian −0.568*** (0.039) −0.533*** (0.040) −0.534*** (0.040)

Black −0.280*** (0.028) −0.242*** (0.031) −0.242*** (0.031)

Latino −0.329*** (0.028) −0.289*** (0.030) −0.289*** (0.030)

Middle Eastern −0.543*** (0.049) −0.491*** (0.051) −0.491*** (0.051)

Native American −0.419*** (0.066) −0.394*** (0.072) −0.394*** (0.072)

White −0.241*** (0.025) −0.182*** (0.027) −0.182*** (0.027)

LSAT score (1 =120–145, 7 =171–180) −0.017*** (0.004) −0.017*** (0.004)

Age (years) −0.001 (0.001) −0.001 (0.001)

Gender (1 =male, 0 =female) −0.005 (0.009) −0.005 (0.009)

International (1 =yes, 0 =no) 0.130*** (0.025) 0.131*** (0.025)

Class year: First (reference category)

Class year: Second 0.0003 (0.010) 0.0002 (0.010)

Class year: Third 0.039*** (0.011) 0.038*** (0.011)

School type: Public (vs. private) 0.033 (0.029)

Enrollment size: <500 students

(reference category)

Enrollment size: 500–900 students −0.045*(0.023)

Enrollment size: >900 students −0.038 (0.034)

AIC 139,308 124,682 124,699

BIC 139,388 124,813 124,856

Log likelihood −69,645 −62,326 −62,331

Note. Unstandardized regression coefﬁcients are displayed, with standard errors in parentheses. LSAT =Law School Admission Test; AIC =Akaike

information criterion; BIC =Bayesian information criterion.

†

p<.10. *p<.05. ** p<.01. *** p<.001.

6LU

Template Version: 11 October 2021 ▪6:01 pm IST PSP-A-2020-1216_format_ﬁnal ▪8 November 2021 ▪3:53 pm IST

These converging results highlight the robustness of EAs’high

ethnic homophily.

Discussion

Analyzing a survey of 54,620 students from 124 U.S. law schools,

Study 1 provided initial evidence for EAs’high ethnic homophily.

The large sample size and the robustness checks underscored this

ﬁnding’s reliability.

Study 2

Study 2 extended Study 1 in three ways. First, whereas Study 1

analyzed a large-scale survey, Study 2 used social network analysis

as a complementary methodology. Whereas Study 1 used a survey

measure of ethnic homophily, Study 2 used a well-established social

network measure of ethnic homophily (point correlation coefﬁ-

cient). Second, Study 2 tested the generalizability of our theoretical

perspective by examining another context: MBA students in a U.S.

business school. Third, whereas Study 1 provided evidence for EAs’

high ethnic homophily (Hypothesis 2), Study 2 tested our whole

theoretical perspective: EAs are less likely than other ethnicities to

emerge as leaders in a multiethnic environment because of higher

ethnic homophily (Hypotheses 1–4).

The research design of Study 2 had notable methodological

strengths. First, MBA students were assigned to different class

sections by school; that is, they did not self-select into their class

sections based on certain criteria (e.g., career interests). Second, all

class sections started without designated leaders, which precluded

the possibility of reverse causality, where being a leader subse-

quently inﬂuences one’s friendship ties. Third, because the students

just entered the MBA program, most of them did not have pre-

existing friendship ties with classmates. This setting enabled us to

examine the natural processes of friendship formation and leader-

ship emergence in a new multiethnic environment (i.e., high eco-

logical validity). Fourth, all students participated in all surveys as a

part of required MBA assignments, which precluded potential self-

selection bias in survey participation. Fifth, our focal variables were

assessed with different data sources at different time points (see

Table 3), which both mitigated common source bias and established

temporal precedence. In particular, as detailed in the Method section

below, we followed the literature (Mollica et al., 2003) and

computed ethnic homophily based on outward friend nominations

(i.e., an individual’s nominations of others as friends), while we

operationalized leadership attainment as the number of inward

leader nominations (i.e., others’nominations of the individual as

leader); this methodological approach mitigates common source

bias. Sixth, we measured assertiveness with peer ratings instead of

self-ratings, precluding the possibility that EAs might rate them-

selves lower on assertiveness due to modesty. Finally, to rule out

alternative explanations, we considered a host of potential con-

founds, including network centrality, English proﬁciency, interna-

tional/American status, SES, personality, and demographics.

Method

Study 2 was conducted at a top U.S. business school. Each year,

the school has two entering cohorts: One cohort enters in August and

graduates after four semesters, whereas the other cohort enters in

January and graduates after three semesters. The two cohorts are

similar, though the January-entry cohort is smaller and more

international.

Sample

Participants (N=202) were a complete January-entry cohort of

full-time MBA students (35.6% female; M

age

at matriculation =

28.20, SD =2.85; 60.4% international). Upon entering the MBA

program, the students were divided into three class sections. Each

section had about 70 students (M=67.33, SD =.58), who took all

core courses together.

Based on self-reported ethnicities, 15% of the students were EA,

10% SA, 47% White, 15% Latino, 2% Black, 4% Middle Eastern/

Arab, and 7% multiethnic/other. Our analyses focused on EAs,

SAs, Latinos, and Whites, because sample sizes of the other ethnic

categories were too small (e.g., some class sections had only one

Black student, so ethnic homophily could not be calculated for

Blacks).

Leader Nominations (Outcome)

After an intensive 1-week MBA orientation, all students com-

pleted a conﬁdential survey as part of a required core class. Each

student was asked to nominate one to ﬁve classmates whom he/she

This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Table 3

Study 2: Research Design

Variable Measurement Source When

Ethnicity Self-reported Required core-class survey 1 week after MBA started

Ethnic homophily Computed based on outward nominations Required core-class survey 1 week after MBA started

Assertiveness Other-rated Required peer evaluation 5 weeks after MBA started

Leader nominations Other-nominated Required core-class survey 1 week after MBA started

Eigenvector centrality Computed Required core-class survey 1 week after MBA started

Big Five personality Self-rated Required self-evaluation 2 weeks after MBA started

Age, gender, international/American status,

socioeconomic status, marital status,

child status

Self-reported Required core-class survey 1 week after MBA started

Whether MBA was sponsored School recorded Career Management Center Pre-MBA

Post-MBA employment country School recorded Career Management Center 4 months after graduation

Note. MBA =Master of Business Administration.

A SOCIAL NETWORK PERSPECTIVE ON THE BAMBOO CEILING 7

Template Version: 11 October 2021 ▪6:01 pm IST PSP-A-2020-1216_format_ﬁnal ▪8 November 2021 ▪3:53 pm IST

viewed as leaders in his/her class section (of about 70 students). By

the time of the survey, the students were already well acquainted

with classmates in their section. To facilitate leader nominations, we

displayed the class section roster alphabetically along with each

student’s proﬁle photo. We tallied the number of times a student was

nominated as a leader.

Friendship Network

Each student also responded to the following question: “Please

indicate 1 to 10 individuals in your class section that you would say

you are friends with at this time. List your closest friend ﬁrst,

followed by your next closest friend, etc. It is OK to list fewer than

10 people.”Students selected their friends from 10 identical drop-

down rosters (i.e., Choice #1, Choice #2, ::: Choice #10). These

questions enabled us to construct the entire friendship network of

each class section. Figure 2 visualizes the friendship network of one

class section.

Ethnic Homophily (Focal Mediator)

Following past studies (Ibarra, 1992;Mehra et al., 1998;Mollica

et al., 2003), we calculated each person’s ethnic homophily based

on which classmates the person nominated as his/her friends

This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Figure 2

The Friendship Network of an MBA Class Section

Note.“EA”=East Asian; “SA”=South Asian; “ME”=Middle Easterner; MBA =Master of Business Administration. See the online article

for the color version of this ﬁgure.

8LU

Template Version: 11 October 2021 ▪6:01 pm IST PSP-A-2020-1216_format_ﬁnal ▪8 November 2021 ▪3:53 pm IST

(i.e., outward friend nominations). The point correlation coefﬁ-

cient (Ibarra, 1992;Krackhardt, 1990;Lawrence & Shah, 2020;

Mehra et al., 1998;Mollica et al., 2003), also known as the phi

coefﬁcient (Gaol et al., 2019), is a well-established measure of

homophily:

ad −bc

ﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ

ða+cÞðb+dÞða+bÞðc+dÞ

where ais the number of same-ethnicity friends the person nomi-

nated, bis the number of cross-ethnicity friends the person nominated,

cis the number of same-ethnicity friends the person could have

nominated but did not, and dis the number of cross-ethnicity friends

the person could have nominated but did not (Mollica et al., 2003,

p. 128). Homophily scores range from −1to+1, with positive values

indicating homophily, negative values indicating heterophily, and

zero indicating perfect neutrality (Lawrence & Shah, 2020).

Importantly, this formula “indicates the proportion of homophily

that is beyond what would be expected if individuals were randomly

paired in a given setting”(Mollica et al., 2003, p. 128). In other

words, the formula accounts for the availability of ethnic ingroup

and outgroup peers in a given class section. This is important

because Whites were the majority, so there were naturally more

Whites available to become friends with.

Assertiveness (Parallel Mediator)

About 5 weeks after the MBA program started, each student was

rated anonymously by at least four classmates as part of a required

peer evaluation. To measure assertiveness, we used the 3-item scale

from Wallen et al. (2017):“X speaks up and shares his/her views

when it is appropriate”;“X is willing to engage in constructive

interpersonal confrontations”;“X is able to stand his/her ground in a

heated conﬂict”(1 =strongly disagree,7=strongly agree;α=

.84). We computed each student’s assertiveness by averaging across

his/her classmates’ratings.

Control Variables

We collected a broad set of control variables to rule out alternative

explanations.

Network Centrality. One alternative explanation is that EAs’

low leadership emergence is actually driven by their low network

centrality rather than by their high ethnic homophily. Indeed, prior

research has found that individuals with higher centrality in a

network are more likely to attain leadership (Neubert & Taggar,

2004). Whereas ethnic homophily captures an individual’s prefer-

ence for socializing with individuals of the same ethnicity, centrality

captures an individual’sgeneral popularity (regardless of ethnicity).

To rule out general popularity as an alternative explanation, we

computed eigenvector centrality, which captures the extent to which

an individual has many friends who themselves have many friends

(Bonacich, 2007).

Eigenvector centrality scores range from 0 to 1, with higher scores

indicating higher centrality. To compute eigenvector centrality for

each student, we used function “eigen_centrality”in R package

igraph. For each person, we computed three versions of eigenvector

centrality based on (a) inward friend nominations, (b) outward friend

nominations, and (c) reciprocal friend nominations, respectively.

Following past studies on eigenvector centrality (Ballinger et al.,

2016;Webster et al., 2016), we present analyses based on inward

friend nominations; results were robust when we used outward or

reciprocal friend nominations instead.

Number of Friend Nominations. We controlled for the num-

ber of friends one nominated because it could be a confounding

variable associated with ethnicity or ethnic homophily (Gibbons &

Olk, 2003). For example, individuals who selected fewer friends

might be more ethnically homophilous. A one-way analysis of

variance (ANOVA) found no signiﬁcant ethnic difference in the

mean number of friend nominations (F=.76, p=.52).

English Proﬁciency. We controlled for English proﬁciency to

rule out the possibility that EAs’high ethnic homophily and low

leadership emergence are driven by low English proﬁciency. We

procured scores of the Verbal section of the required MBA admission

exams: the Graduate Management Admission Test (GMAT) and the

Graduate Record Examinations (GRE).

About 90% of students took

the GMAT and about 10% took the GRE. To standardize across the

two exams in different years, we followed the ofﬁcial GMAT and

GRE websites to convert all scores to percentiles.

Post-MBA Employment Country. It is possible that EAs

were more inclined to work outside the United States (e.g., in

EA countries) after completing their MBA and thus more moti-

vated to develop friendships with other EAs (rather than ethnic

outgroup classmates). To rule out this possibility, we procured

post-MBA employment information from the business school’s

Career Management Center. Following the guidelines of the

MBA Career Services & Employer Alliance (www.mbacsea

.org), the business school records the ﬁrst post-MBA employment

of each student 4 months after graduation. Analyses found that

EAs and non-EAs did not differ signiﬁcantly in their likelihood to

work in the United States after earning their MBA (χ

=.09,

p=.76); this result suggests that post-MBA employment country

was unlikely a confounding variable.

MBA Sponsorship. It is possible that EAs were more likely to

be sponsored by organizations for their MBA degree and thus less

motivated to network with ethnic outgroup classmates. To rule out

this possibility, we procured MBA sponsorship data from the Career

Management Center. EAs and non-EAs did not differ signiﬁcantly

in MBA sponsorship (χ

=.38, p=.54), suggesting that it was

unlikely a confounding variable.

Marital Status and Child Status. It is possible that EAs were

more likely to have spouses or children and thus spend less time

socializing with ethnic outgroup classmates. Of the 202 students,

6.4% had children. About 25.2% were married, 23.3% were in

committed relationships, and 51.5% were single. EAs and non-

EAs did not differ signiﬁcantly in child status (χ

=.01, p=.93)

or marital status (χ

=.25, p=.62), suggesting that they were

unlikely confounding variables.

Socioeconomic Status. We controlled for SES because it could

be a confounding variable associated with ethnicity, ethnic homo-

phily, or leadership emergence (Lu et al., 2020). At the end of the

social network survey, we measured SES with the widely used

“ladder question”(Adler et al., 2000), which featured a drawing of a

10-rung ladder representing all of the people in the United States and

This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Research found that for non-native English speakers who took both GRE

Verbal and Test of English as a Foreign Language (TOEFL), the two scores

were highly correlated at r=.82 (Pesta et al., 2019).

A SOCIAL NETWORK PERSPECTIVE ON THE BAMBOO CEILING 9

Template Version: 11 October 2021 ▪6:01 pm IST PSP-A-2020-1216_format_ﬁnal ▪8 November 2021 ▪3:53 pm IST

asked students to place themselves on the ladder in terms of SES

(1 =lowest, 10 =highest).

Personality. We controlled for the Big Five personality traits

(openness to experience, conscientiousness, extraversion, agreeable-

ness, and emotional stability) because they could be confounding

variables associated with ethnicity, ethnic homophily, or leadership

emergence (Bendersky & Shah, 2013;Feiler & Kleinbaum, 2015;

Judge et al., 2002). To assess the Big Five, we adopted the widely

used Ten-Item Personality Inventory (Gosling et al., 2003; Lu, Liu,

Liao, & Wang, 2020), which the students completed as part of a

required self-evaluation (1 =strongly disagree,7=strongly agree)

2 weeks after the MBA program started.

Other Demographics. Finally, we controlled for age, gender,

and whether a student was international.

Results

Descriptive statistics and Pearson correlations are displayed in

Table 4. Because students were nested within class sections, we

conducted multilevel regressions to account for within-section

statistical dependence.

East Asians →Higher Ethnic Homophily

(Focal Mediator)

Consistent with prior research (McPherson et al., 2001), mean

ethnic homophily was positive for all the four ethnicities (M=.12).

This result indicates that each ethnicity exhibited homophily to

some extent.

As detailed in Table 5 Model 1, consistent with Hypothesis 2,

EAs (M=.24) were more homophilous than SAs (M=.11,

B=.13, SE =.05, p=.006), Latinos (M=.15, B=.09,

SE =.04, p=.030), and Whites (M=.07, B=.17, SE =.03,

p<.001). By contrast, SAs did not differ signiﬁcantly from

Latinos (B=−.03, SE =.05, p=.48) or Whites (B=.04,

SE =.04, p=.24). These results were substantively similar after

we accounted for the control variables in multilevel linear regres-

sions (Table 5 Models 2–4). As explained in the Method section,

the measure of ethnic homophily (point correlation coefﬁcient)

already accounted for the ethnic groups’differential sizes in a

given class section.

East Asians →Lower Assertiveness

(Parallel Mediator)

Consistent with prior research (Lu et al., 2020), EAs (M=5.18)

were signiﬁcantly less assertive than SAs (M=5.62, B=−.44,

SE =.15, p=.003), Latinos (M=5.72, B=−.54, SE =.14,

p<.001), and Whites (M=5.77, B=−.59, SE =.11, p<.001).

By contrast, SAs did not differ signiﬁcantly from Latinos (B=−.10,

SE =.12, p=.42) or Whites (B=−.15, SE =.13, p=.26). These

results were robust after we accounted for the control variables.

East Asians →Lower Leadership Emergence

For the outcome variable of leader nominations, we conducted

multilevel Poisson regressions because leader nomination was a

positively skewed, count variable that took only nonnegative integer

values. Consistent with Hypothesis 3, EAs (M=.39) were

signiﬁcantly less likely to be nominated as leaders than SAs

(M=2.81, B=−1.97, SE =.32, p<.001), Latinos (M=3.30,

B=−2.14, SE =.31, p<.001), and Whites (M=4.46,

B=−2.45, SE =.29, p<.001). These results were robust after

we accounted for the control variables.

Ethnic Homophily →Lower Leadership Emergence

Consistent with Hypothesis 1, ethnic homophily negatively pre-

dicted leader nominations in a multilevel Poisson regression

(Table 6 Model 1: B=−2.50, SE =.26, p<.001). This effect

was robust (Table 6 Model 2: B=−2.32, SE =.27, p<.001) after

we controlled for assertiveness, while assertiveness positively pre-

dicted leader nominations (Table 6 Model 2: B=.98, SE =.09,

p<.001). The effects of ethnic homophily and assertiveness were

both robust after we further accounted for the control variables,

including eigenvector centrality, English proﬁciency, demographics,

and personality (Table 6 Models 3 and 4).

Mediation Analyses

Consistent with Hypothesis 4, ethnic homophily signiﬁcantly

mediated the effect of ethnicity (1 =EA, 0 =SA/Latino/White)

on leader nominations, indirect effect =−.82, bias-corrected boot-

strapped 95% CI [−1.05, −.47]. Consistent with prior research

(Lu et al., 2020), assertiveness also signiﬁcantlymediatedthe

effect of ethnicity (1 =EA, 0 =SA/Latino/White) on leader

nominations, indirect effect =−1.29, bias-corrected bootstrapped

95% CI [−2.59, −.86].

Next, we conducted simultaneous mediation analysis with ethnic

homophily and assertiveness as parallel mediators (Preacher &

Hayes, 2008). Both ethnic homophily (indirect effect =−.80,

bias-corrected bootstrapped 95% CI [−2.73, −.04]) and assertive-

ness (indirect effect =−1.24, bias-corrected bootstrapped 95% CI

[−2.65, −.47]) were signiﬁcant mediators. These results suggest that

ethnic homophily explained EAs’lower leadership emergence

above and beyond assertiveness (Figure 3).

Discussion

By conducting social network analysis on three complete class

sections of new MBA students, Study 2 supported our theoretical

perspective. First, while all four ethnicities exhibited ethnic

This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

One potential limitation is that the Ten-Item Personality Inventory (TIPI)

uses only two items to measure each of the Big Five personality traits.

Unfortunately, we were unable to use a longer personality scale because of

survey length limit. Nevertheless, the TIPI had satisfactory psychometric

properties in both Studies 2 and 3. Consider the case of extraversion. First, its

two items (“extraverted, enthusiastic”;“reserved, quiet”) were highly corre-

lated (r=−.70, p<.001). Second, extraversion was positively correlated

with assertiveness and leadership emergence, which is consistent with the

literature (Bendersky & Shah, 2013;Judge et al., 2002). Third, extraversion

was positively correlated with openness to experience, but not signiﬁcantly

correlated with agreeableness, conscientiousness, or emotional stability;

meanwhile, conscientiousness was positively and signiﬁcantly correlated

with agreeableness and emotional stability. These results are consistent with

the well-established “Big Two”pattern in the personality literature, where

agreeableness, conscientiousness, and emotional stability represent a higher-

order factor “Alpha,”and extraversion and openness represent a higher-order

factor “Beta”(DeYoung, 2006;Digman, 1997;Wei et al., 2017).

10 LU

Template Version: 11 October 2021 ▪6:01 pm IST PSP-A-2020-1216_format_ﬁnal ▪8 November 2021 ▪3:53 pm IST

This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Table 4

Study 2: Descriptive Statistics and Pearson Correlations

Variable MSD 12345678910111213141516171819

1. East Asian (vs. non-East Asian) 0.18 0.38 —

2. Leader nominations received 3.43 6.73 −.21** —

3. Ethnic homophily 0.12 0.18 .32** −.20** —

4. Assertiveness 5.63 0.58 −.37** .23** −.15*—

5. Eigenvector centrality 0.67 0.24 −.07 .08 .15 .06 —

6. Age 28.20 2.85 .21** .12 .17*−.12 .02 —

7. Gender (1 =male, 0 =female) 0.64 0.48 −.19*.12 −.05 .23** .09 .07 —

8. International (1 =yes, 0 =no) 0.60 0.49 .17*−.22** .23** −.07 −.01 −.02 −.05 —

9. GMAT/GRE Verbal percentile 87.27 10.55 −.14 .10 −.03 .19** .04 −.08 .26** −.14*—

10. Socioeconomic status 8.11 1.34 −.17*.05 .00 .20** −.02 −.06 .13 −.01 .22** —

11. Number of friend nominations 8.00 2.60 −.03 .08 .14 .06 .92** .07 .13 .03 .06 .00 —

12. Married (1 =yes, 0 =no) 0.25 0.44 −.04 .12 .06 .00 −.03 .36** .03 −.04 −.07 .06 −.03 —

13. Has children (1 =yes, 0 =no) 0.06 0.25 −.01 .01 .09 .02 .10 .34** −.02 .01 .02 −.01 .08 .40** —

14. MBA sponsored (1 =yes, 0 =no) 0.24 0.43 −.05 −.01 .07 .19** −.06 −.10 .09 .30** −.07 .05 −.05 −.09 −.01 —

15. Post-MBA job in U.S. (1 =yes, 0 =no) 0.58 0.50 −.02 .10 −.17*.08 .04 .11 .00 −.62** .07 .00 .01 .07 .06 −.48** —

16. Openness to experience 5.29 1.07 −.12 .19** −.21** .15*.13 −.05 −.05 −.06 −.04 −.11 .12 −.12 −.05 .04 .02 —

17. Conscientiousness 5.43 1.26 −.03 .02 −.03 −.17*−.03 −.15*−.13 −.15*−.07 .05 −.06 .12 −.09 −.10 .09 −.07 —

18. Extraversion 4.62 1.62 −.16*.40** −.14 .36** .19** −.02 −.15*.05 −.01 −.02 .16*.03 .02 .21** −.03 .30** −.06 —

19. Agreeableness 5.14 1.13 .03 −.09 −.04 −.16*−.06 .04 −.13 −.08 .01 .00 −.03 −.07 −.05 −.11 .07 −.04 .20** .01 —

20. Emotional stability 5.09 1.23 .05 .01 .06 .09 .07 −.02 .12 −.08 .02 −.02 .07 .15*.05 −.05 .10 .09 .26** .06 .13

Note.“non-East Asian”=South Asian, Latino, and White. GMAT =Graduate Management Admission Test; GRE =Graduate Record Examinations; MBA =Master of Business Administration.

*p<.05. ** p<.01.

A SOCIAL NETWORK PERSPECTIVE ON THE BAMBOO CEILING 11

Template Version: 11 October 2021 ▪6:01 pm IST PSP-A-2020-1216_format_ﬁnal ▪8 November 2021 ▪3:53 pm IST

homophily, EAs were signiﬁcantly more ethnically homophilous

than SAs, Latinos, and Whites; this effect was robust to a host

of control variables (e.g., network centrality, English proﬁ-

ciency, SES, personality). Second, consistent with recent research

(Lu et al., 2020), EAs were signiﬁcantly less assertive than SAs,

Latinos, and Whites. Third, ethnic homophily partly explained why

EAs (but not SAs) were less likely to be nominated as leaders—above

and beyond the mechanism of assertiveness. Together, these

This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Table 5

Study 2: Multilevel Linear Regressions Predicting Ethnic Homophily

Variable

Model 1 Model 2 Model 3 Model 4

BSEBSEBSEBSE

East Asian (reference category)

South Asian −0.128** (0.046) −0.132** (0.047) −0.143** (0.051) −0.111*(0.051)

Latino −0.091*(0.042) −0.088*(0.042) −0.085

†

(0.045) −0.074

†

(0.044)

White −0.174*** (0.034) −0.150*** (0.038) −0.160*** (0.040) −0.137*** (0.039)

Assertiveness −0.012 (0.023) −0.014 (0.025) 0.008 (0.026)

Age (years) 0.005 (0.004) 0.004 (0.005) 0.005 (0.005)

Gender (1 =male, 0 =female) 0.001 (0.027) −0.007 (0.028) −0.036 (0.029)

International (1 =yes, 0 =no) 0.054*(0.026) 0.017 (0.034) 0.016 (0.034)

GMAT/GRE Verbal percentile 0.001 (0.001) 0.001 (0.001) 0.001 (0.001)

Socioeconomic status 0.009 (0.010) 0.008 (0.010) 0.001 (0.010)

Number of friend nominations 0.012*(0.005) 0.013*(0.005) 0.017** (0.005)

Married (1 =yes, 0 =no) 0.015 (0.034) −0.002 (0.034)

Has children (1 =yes, 0 =no) 0.006 (0.056) 0.005 (0.055)

MBA sponsored (1 =yes, 0 =no) 0.026 (0.036) 0.033 (0.035)

Post-MBA job in U.S. (1 =yes, 0 =no) −0.027 (0.036) −0.030 (0.035)

Openness to experience −0.027*(0.012)

Conscientiousness 0.002 (0.011)

Extraversion −0.016

†

(0.009)

Agreeableness −0.017 (0.012)

Emotional stability 0.014 (0.011)

AIC −121 −119 −102 −104

BIC −102 −78 −49 −36

Log likelihood 66 73 68 74

Note. Unstandardized regression coefﬁcients are displayed, with standard errors in parentheses. GMAT =Graduate Management Admission Test; GRE =

Graduate Record Examinations; MBA =Master of Business Administration; AIC =Akaike information criterion; BIC =Bayesian information criterion.

†

p<.10. *p<.05. ** p<.01. *** p<.001.

Table 6

Study 2: Multilevel Poisson Regressions Predicting Leader Nominations Received

Variable

Model 1 Model 2 Model 3 Model 4

BSEBSEBSEBSE

Ethnic homophily −2.499*** (0.264) −2.318*** (0.275) −2.206*** (0.279) −0.807** (0.291)

Assertiveness 0.984*** (0.088) 0.842*** (0.094) 0.406*** (0.100)

Eigenvector centrality 1.164*** (0.232) −0.403

†

(0.234)

Age (years) 0.104*** (0.013) 0.071*** (0.014)

Gender (1 =male, 0 =female) 0.381*** (0.113) 0.716*** (0.118)

International (1 =yes, 0 =no) −0.601*** (0.088) −0.852*** (0.093)

GMAT/GRE Verbal percentile 0.014** (0.005) 0.008 (0.006)

Socioeconomic status 0.001 (0.038) 0.106*(0.044)

Openness to experience 0.305*** (0.053)

Conscientiousness 0.197*** (0.042)

Extraversion 0.559*** (0.038)

Agreeableness −0.156*** (0.047)

Emotional stability 0.045 (0.040)

AIC 1,583 1,446 1,265 916

BIC 1,592 1,458 1,296 964

Log likelihood −788 −719 −622 −443

Note. Unstandardized regression coefﬁcients are displayed, with standard errors in parentheses. GMAT =Graduate Management Admission Test; GRE =

Graduate Record Examinations; AIC =Akaike information criterion; BIC =Bayesian information criterion.

†

p<.10. *p<.05. ** p<.01. *** p<.001.

12 LU

Template Version: 11 October 2021 ▪6:01 pm IST PSP-A-2020-1216_format_ﬁnal ▪8 November 2021 ▪3:53 pm IST

ﬁndings suggest that high ethnic homophily and low assertiveness

both account for EAs’disadvantage in leadership emergence.

Study 3

Study 3 extended Study 2 in three ways. First, Study 3 examined a

sample twice as large to ascertain the replicability of Study 2’s

ﬁndings. Second, whereas Study 2 measured leader nominations,

Study 3 also measured whether a person was actually elected as a

class-section leader. This is a consequential outcome because many

MBAs list their leadership positions on résumés and LinkedIn

proﬁles, as many employers favor students with such leadership

experiences. Third, the leader election outcome was measured 1 year

after we measured ethnic homophily; this time-lagged design

established temporal precedence.

Method

Sample

Participants (N=560) were a complete August-entry cohort of

full-time MBA students of the same business school (40.0% female;

age

at matriculation =27.97, SD =2.48; 41.3% international). As

in Study 2, the students were divided into class sections of about 70

students (M=70.00, SD =1.60) upon entering the MBA program.

Whereas Study 2 involved three class sections, Study 3 involved

eight class sections.

Based on self-reported ethnicities, 21% of the students were EA,

7% SA, 51% White, 9% Latino, 4% Black, 3% Middle Eastern/

Arab, and 5% multiethnic/other. As in Study 2, our analyses focused

on EAs, SAs, Latinos, and Whites, because sample sizes of the other

ethnic categories were too small.

Leader Nominations

As in Study 2, after an intensive 1-week MBA orientation, all

students completed a conﬁdential survey as part of a required core

class. Each student was asked to nominate one to ﬁve classmates

whom he/she viewed as leaders. To facilitate leader nominations, we

displayed the class section roster alphabetically along with each

student’s proﬁle photo.

Leader Election (1 Year Later)

In addition to leader nominations, we also obtained data from

the Ofﬁce of Student Affairs on whether a student was elected as a

class-section leader 1 year later (1 =yes, 0 =no). At the beginning

of Year 2 of the MBA program, each class section held a competitive

election for its leadership positions (e.g., class-section chair, social

chair, academic chair, career chair). The MBA program values

student governance, so these leaders are responsible for class-

section activities. The elected leaders facilitate class-section unity,

manage class-section funds ($4,000), act as liaisons to the adminis-

tration and faculty, and serve as leaders during graduation events, in

addition to handling other responsibilities. Because class-section

leaders have a large impact on peers who vote for them, these

elections carry high stakes.

Mediators and Control Variables

Ethnic homophily (focal mediator), assertiveness (parallel medi-

ator), and control variables (eigenvector centrality, number of friend

nominations, English proﬁciency, post-MBA employment country,

MBA sponsorship, marital status, child status, SES, the Big Five

personality traits, international/American status, age, gender) were

measured in the same way as in Study 2. As detailed in Table 7,

these variables were assessed with different data sources at different

time points, which mitigated common source bias and established

temporal precedence.

Results

Descriptive statistics and Pearson correlations are displayed in

Table 8. Because MBA students were nested within class sections,

we conducted multilevel regressions to account for within-section

statistical dependence.

East Asians →Higher Ethnic Homophily

(Focal Mediator)

As detailed in Table 9 Model 1, consistent with Hypothesis 2 and

Study 2, EAs (M=.20) were signiﬁcantly more ethnically homo-

philous than SAs (M=.12, B=.08, SE =.03, p=.005), Latinos

(M=.14, B=.06, SE =.03, p=.018), and Whites (M=.06,

This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Figure 3

Study 2: Simultaneous Mediation Analysis for Leader Nominations

A SOCIAL NETWORK PERSPECTIVE ON THE BAMBOO CEILING 13

Template Version: 11 October 2021 ▪6:01 pm IST PSP-A-2020-1216_format_ﬁnal ▪8 November 2021 ▪3:53 pm IST

B=.14, SE =.02, p<.001). These results were robust after we

accounted for the control variables in multilevel linear regressions

(Table 9 Models 2–4).

East Asians →Lower Assertiveness (Parallel Mediator)

Consistent with Study 2, EAs (M=5.33) were signiﬁcantly less

assertive than SAs (M=5.62, B=−.29, SE =.10, p=.003),

Latinos (M=5.79, B=−.45, SE =.09, p<.001), and Whites

(M=5.72, B=−.39, SE =.06, p<.001). These results were

robust after we accounted for the control variables.

East Asians →Lower Leadership Emergence

For the outcome variable of leader nominations, we conducted

multilevel Poisson regressions because leader nomination was a

positively skewed, count variable that took only nonnegative integer

values. Consistent with Hypothesis 3 and Study 2, EAs (M=.85)

were signiﬁcantly less likely to be nominated as leaders than SAs

(M=4.87, B=−1.74, SE =.12, p<.001), Latinos (M=3.39,

B=−1.39, SE =.13, p<.001), and Whites (M=4.09, B=

−1.60, SE =.10, p<.001). These results were robust after we

accounted for the control variables.

In further support of Hypothesis 3, EAs (8.5%) were less likely

than SAs (25.6%; χ

=7.65, p=.006), Latinos (20.4%; χ

=4.58,

p=.032), and Whites (16.7%; χ

=4.52, p=.033) to be elected as

class-section leaders. These results were substantively similar in

multilevel logistic regressions that accounted for the control variables.

Ethnic Homophily →Lower Leadership Emergence

Consistent with Hypothesis 1 and Study 2, ethnic homophily

negatively predicted leader nominations in a multilevel Poisson

regression (Table 10 Model 1: B=−2.21, SE =.16, p<.001).

This effect was robust (Table 10 Model 2: B=−2.15, SE =.17,

p<.001) after we controlled for assertiveness, while assertiveness

positively predicted leader nominations (Table 10 Model 2:

B=1.00, SE =.06, p<.001). The effects of ethnic homophily

and assertiveness were both robust after we further accounted for the

control variables, including eigenvector centrality, English proﬁ-

ciency, demographics, and personality (Table 10 Models 3 and 4).

In further support of Hypothesis 1, ethnic homophily negatively

predicted leader election in a multilevel logistic regression (Table 11

Model 1: B=−2.77, SE =.83, Wald z =−3.34, p<.001). This

effect was robust (Table 11 Model 2: B=−2.63, SE =.86, Wald

z=−3.07, p=.002) after we controlled for assertiveness, while

assertiveness positively predicted leader election (Table 11 Model 2:

B=.71, SE =.26, Wald z =2.70, p=.007). The effects of ethnic

homophily and assertiveness were robust after we further accounted

for the control variables (Table 11 Models 3 and 4).

Mediation Analyses

Consistent with Hypothesis 4 and Study 2, ethnic homophily

signiﬁcantly mediated the effect of ethnicity (1 =EA, 0 =SA/

Latino/White) on leader nominations, indirect effect =−.54, bias-

corrected bootstrapped 95% CI [−.98, −.19]. Consistent with Study

2, assertiveness also signiﬁcantly mediated the effect of ethnicity

(1 =EA, 0 =SA/Latino/White) on leader nominations, indirect

effect =−.77, bias-corrected bootstrapped 95% CI [−1.14, −.39].

Next, we conducted simultaneous mediation analysis with ethnic

homophily and assertiveness as parallel mediators (Preacher & Hayes,

2008). Consistent with Study 2, both ethnic homophily (indirect

effect =−.54, bias-corrected bootstrapped 95% CI [−1.04, −.20])

and assertiveness (indirect effect =−.77, bias-corrected boot-

strapped 95% CI [−1.29, −.43]) were signiﬁcant mediators

(Figure 4).

In further support of Hypothesis 4, ethnic homophily signiﬁcantly

mediated the effect of ethnicity (1 =EA, 0 =SA/Latino/White) on

leader election, indirect effect =−.04, bias-corrected bootstrapped

95% CI [−.05, −.02]. Assertiveness also signiﬁcantly mediated the

effect of ethnicity (1 =EA, 0 =SA/Latino/White) on leader elec-

tion, indirect effect =−.03, bias-corrected bootstrapped 95% CI

[−.04, −.02].

Next, we conducted simultaneous mediation analysis with ethnic

homophily and assertiveness as parallel mediators (Preacher & Hayes,

2008). Both ethnic homophily (indirect effect =−.03, bias-corrected

bootstrapped 95% CI [−.06, −.01]) and assertiveness (indirect

effect =−.03, bias-corrected bootstrapped 95% CI [−.05, −.01])

were signiﬁcant mediators (Figure 5).

This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Table 7

Study 3: Research Design

Variable Measurement Source When

Ethnicity Self-reported Required core-class survey 1 week after MBA started

Ethnic homophily Computed based on outward nominations Required core-class survey 1 week after MBA started

Assertiveness Other-rated Required peer evaluation 5 weeks after MBA started

Leader nominations Other-nominated Required core-class survey 1 week after MBA started

Whether elected as leader School recorded Ofﬁce of Student Affairs 1 year after MBA started

Eigenvector centrality Computed Required core-class survey 1 week after MBA started

Big Five personality Self-rated Required self-evaluation 2 weeks after MBA started

Age, gender, international/American status,

socioeconomic status, marital status,

child status

Self-reported Required core-class survey 1 week after MBA started

Whether MBA was sponsored School recorded Career Management Center Pre-MBA

Post-MBA employment country School recorded Career Management Center 4 months after graduation

Note. MBA =Master of Business Administration.

14 LU

Template Version: 11 October 2021 ▪6:01 pm IST PSP-A-2020-1216_format_ﬁnal ▪8 November 2021 ▪3:53 pm IST

This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Table 8

Study 3: Descriptive Statistics and Pearson Correlations

Variable MSD 1234567891011121314151617181920

1. East Asian (vs. non-East Asian) 0.24 0.43 —

2. Leader nominations received 3.65 7.90 −.19** —

3. Elected as leader (1 =yes, 0 =no) 0.16 0.36 −.11*.29** —

4. Ethnic homophily 0.10 0.17 .32** −.15** −.15** —

5. Assertiveness 5.63 0.56 −.30** .21** .16** −.13** —

6. Eigenvector centrality 0.65 0.25 −.04 .12** .12** −.05 .05 —

7. Age 27.97 2.48 .14** .05 .00 .11*−.05 −.03 —

8. Gender (1 =male, 0 =female) 0.60 0.49 −.10*.02 −.01 .04 .05 .13** .17** —

9. International (1 =yes, 0 =no) 0.41 0.49 .29** −.15** −.07 .22** −.12** −.05 .12** .05 —

10. GMAT/GRE Verbal percentile 90.56 8.80 −.18** −.02 .01 −.14** .04 .01 −.16** .05 −.22** —

11. Socioeconomic status 8.23 1.40 −.18** −.08 .00 −.03 .08 .00 −.14** .12** −.13** .17** —

12. Number of friend nominations 8.17 2.64 .03 .11** .13** −.02 .02 .89** .00 .14** .01 .02 .01 —

13. Married (1 =yes, 0 =no) 0.21 0.41 .08 .06 .04 .06 .02 .01 .31** .11*.13** −.13** −.03 .03 —

14. Has children (1 =yes, 0 =no) 0.04 0.19 .08 .05 .02 −.01 .06 −.01 .27** .08 .07 −.01 −.13** −.01 .35** —

15. MBA sponsored (1 =yes, 0 =no) 0.07 0.25 −.03 .03 .08 −.09 .03 .01 −.05 .09*.16** −.03 −.06 .00 .02 .06 —

16. Post-MBA job in U.S. (1 =yes, 0 =no) 0.80 0.40 −.26** .07 −.01 −.11*.10*.04 −.10*−.05 −.52** .14** .04 −.03 −.02 −.08 −.24** —

17. Openness to experience 5.21 1.12 −.13** .09*.02 −.13** .05 −.02 .05 −.07 .07 −.08 .04 −.02 −.06 −.04 −.01 .02 —

18. Conscientiousness 5.65 1.08 −.10*.01 .00 −.02 −.05 −.04 −.01 −.14** .01 −.05 −.09*−.06 −.06 −.03 −.03 .11** .08 —

19. Extraversion 4.64 1.59 −.17** .31** .18** −.09 .26** .23** −.01 −.04 −.01 −.08 .02 .20** .03 −.04 .00 .09*.27** .02 —

20. Agreeableness 5.03 1.16 .06 .08 .05 −.10*−.22** −.12** −.05 −.24** −.07 −.07 −.04 −.10*.01 −.03 −.04 .06 .14** .12** .06 —

21. Emotional stability 5.03 1.28 −.03 .02 .04 −.02 .03 .05 .04 .20** .01 −.07 .03 .05 .09*.03 .08 .09*.12** .13** .02 .13**

Note.“non-East Asian”=South Asian, Latino, and White. GMAT =Graduate Management Admission Test; GRE =Graduate Record Examinations; MBA =Master of Business Administration.

*p<.05. ** p<.01.

A SOCIAL NETWORK PERSPECTIVE ON THE BAMBOO CEILING 15

Template Version: 11 October 2021 ▪6:01 pm IST PSP-A-2020-1216_format_ﬁnal ▪8 November 2021 ▪3:53 pm IST

Discussion

Supporting our theoretical perspective, Study 3 replicated and

extended Study 2’sﬁndings by examining another eight class

sections of MBA students. EAs were less likely than SAs, Latinos,

and Whites to be (a) nominated as leaders and (b) elected as leaders.

EAs’high ethnic homophily partly explained EAs’low leadership

emergence—above and beyond the mechanism of assertiveness.

This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Table 9

Study 3: Multilevel Linear Regressions Predicting Ethnic Homophily

Variable

Model 1 Model 2 Model 3 Model 4

BSEBSEBSEBSE

East Asian (reference category)

South Asian −0.084** (0.030) −0.068*(0.030) −0.076*(0.030) −0.061*(0.030)

Latino −0.065*(0.027) −0.070*(0.028) −0.073*(0.029) −0.073*(0.029)

White −0.145*** (0.018) −0.125*** (0.020) −0.127*** (0.020) −0.125*** (0.021)

Assertiveness −0.013 (0.014) −0.012 (0.014) −0.019 (0.014)

Age (years) 0.003 (0.003) 0.002 (0.004) 0.003 (0.004)

Gender (1 =male, 0 =female) 0.022 (0.016) 0.026†(0.016) 0.014 (0.017)

International (1 =yes, 0 =no) 0.023 (0.017) 0.032 (0.019) 0.032 (0.019)

GMAT/GRE Verbal percentile −0.002†(0.001) −0.002†(0.001) −0.002*(0.001)

Socioeconomic status 0.007 (0.006) 0.006 (0.006) 0.008 (0.006)

Number of friend nominations −0.002 (0.003) −0.002 (0.003) −0.002 (0.003)

Married (1 =yes, 0 =no) 0.009 (0.020) 0.009 (0.020)

Has children (1 =yes, 0 =no) −0.031 (0.045) −0.038 (0.044)

MBA sponsored (1 =yes, 0 =no) −0.058†(0.031) −0.054†(0.030)

Post-MBA job in U.S. (1 =yes, 0 =no) 0.016 (0.022) 0.019 (0.022)

Openness to experience −0.016*(0.007)

Conscientiousness 0.003 (0.007)

Extraversion 0.001 (0.005)

Agreeableness −0.017*(0.007)

Emotional stability −0.001 (0.006)

AIC −385 −373 −368 −371

BIC −359 −319 −297 −279

Log likelihood 198 200 201 207

Note. Unstandardized regression coefﬁcients are displayed, with standard errors in parentheses. GMAT =Graduate Management Admission Test; GRE =

Graduate Record Examinations; MBA =Master of Business Administration; AIC =Akaike information criterion; BIC =Bayesian information criterion.

†

p<.10. *p<.05. ** p<.01. *** p<.001.

Table 10

Study 3: Multilevel Poisson Regressions Predicting Leader Nominations Received

Variable

Model 1 Model 2 Model 3 Model 4

B SE B SE B SE B SE

Ethnic homophily −2.210*** (0.164) −2.147*** (0.172) −2.031*** (0.181) −1.749*** (0.185)

Assertiveness 0.998*** (0.057) 1.059*** (0.061) 0.884*** (0.063)

Eigenvector centrality 0.935*** (0.135) 0.309*(0.135)

Age (years) 0.077*** (0.011) 0.064*** (0.012)

Gender (1 =male, 0 =female) −0.234*** (0.055) −0.028 (0.060)

International (1 =yes, 0 =no) −0.438*** (0.059) −0.425*** (0.061)

GMAT/GRE Verbal percentile −0.001 (0.004) 0.006 (0.004)

Socioeconomic status −0.080*** (0.019) −0.065** (0.021)

Openness to experience 0.013 (0.026)

Conscientiousness 0.138*** (0.027)

Extraversion 0.464*** (0.022)

Agreeableness 0.179*** (0.025)

Emotional stability −0.011 (0.022)

AIC 4,859 4,472 4,059 3,375

BIC 4,872 4,489 4,101 3,438

Log likelihood −2,427 −2,232 −2020 −1,673

Note. Unstandardized regression coefﬁcients are displayed, with standard errors in parentheses. GMAT =Graduate Management Admission Test; GRE =

Graduate Record Examinations; AIC =Akaike information criterion; BIC =Bayesian information criterion.

†

p<.10. *p<.05. ** p<.01. *** p<.001.

16 LU

Template Version: 11 October 2021 ▪6:01 pm IST PSP-A-2020-1216_format_ﬁnal ▪8 November 2021 ▪3:53 pm IST

Further Evidence for the Ethnic Homophily

Perspective on the Bamboo Ceiling

The Hypothesized Effects Exist for Both

Internationals and Americans

We conducted further analyses to examine whether the hypothe-

sized effects existed for both (a) international students and (b)

American students. To maximize statistical power, we combined

samples from Studies 2 and 3.

Internationals

In support of Hypothesis 2, EA internationals were signiﬁcantly

higher on ethnic homophily than SA, Latino, and White interna-

tionals (without controls: B=.15, SE =.02, p<.001; with con-

trols: B=.13, SE =.02, p<.001). In support of Hypothesis 3,

EA internationals were signiﬁcantly less likely to be nominated

as leaders than non-EA internationals (without controls: B=−1.73,

SE =.14, p<.001; with controls: B=−1.36, SE =.15, p<

.001). In support of Hypothesis 4, ethnic homophily signiﬁcantly

mediated the effect of ethnicity (1 =EA international, 0 =SA/

Latino/White international) on leader nominations, indirect effect =

−.62, bias-corrected bootstrapped 95% CI [−1.26, −.18].

Americans

In support of Hypothesis 2, EA Americans were signiﬁcantly

higher on ethnic homophily than SA, Latino, and White Americans

(without controls: B=.06, SE =.02, p=.013; with controls:

B=.05, SE =.02, p=.041). In support of Hypothesis 3, EA

Americans were signiﬁcantly less likely to be nominated as leaders

than non-EA Americans (without controls: B=−1.37, SE =.14,

p<.001; with controls: B=−.71, SE =.15, p<.001). In support

of Hypothesis 4, ethnic homophily signiﬁcantly mediated the effect

of ethnicity (1 =EA American, 0 =SA/Latino/White American)

This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Table 11

Study 3: Multilevel Logistic Regressions Predicting Whether a Student Was Elected as a Class-Section Leader

Variable

Model 1 Model 2 Model 3 Model 4

B SEBSEBSEBSE

Ethnic homophily −2.769*** (0.830) −2.627** (0.856) −2.549** (0.881) −2.432** (0.906)

Assertiveness 0.707** (0.262) 0.683*(0.270) 0.624*(0.292)

Eigenvector centrality 1.366*(0.625) 1.142

†

(0.639)

Age (years) 0.095

†

(0.058) 0.106

†

(0.061)

Gender (1 =male, 0 =female) −0.291 (0.275) −0.243 (0.299)

International (1 =yes, 0 =no) −0.040 (0.282) −0.022 (0.289)

GMAT/GRE Verbal percentile 0.022 (0.020) 0.027 (0.020)

Socioeconomic status 0.032 (0.100) 0.024 (0.107)

Openness to experience −0.104 (0.129)

Conscientiousness −0.058 (0.128)

Extraversion 0.251** (0.097)

Agreeableness 0.178 (0.125)

Emotional stability 0.115 (0.110)

AIC 421 413 411 409

BIC 433 430 453 472

Log likelihood −207 −203 −196 −190

Note. Unstandardized regression coefﬁcients are displayed, with standard errors in parentheses. GMAT =Graduate Management Admission Test; GRE =

Graduate Record Examinations; AIC =Akaike information criterion; BIC =Bayesian information criterion.

†

p<.10. *p<.05. ** p<.01. *** p<.001.

Figure 4

Study 3: Simultaneous Mediation Analysis for Leader Nominations

A SOCIAL NETWORK PERSPECTIVE ON THE BAMBOO CEILING 17

Template Version: 11 October 2021 ▪6:01 pm IST PSP-A-2020-1216_format_ﬁnal ▪8 November 2021 ▪3:53 pm IST

on leader nominations, indirect effect =−.25, bias-corrected boot-

strapped 95% CI [−.77, −.03].

Together, these results suggest that our theoretical perspective

applies to both EA internationals and EA Americans: Both were

more homophilous than their non-EA counterparts and thus less

likely to emerge as leaders in multiethnic environments.

Notably, there was also a signiﬁcant interaction between EA/non-

EA and international/American status on ethnic homophily, such

that EA internationals were especially homophilous (without con-

trols: B=.09, SE =.03, p=.004; with controls: B=.09, SE =

.03, p=.009). This result points to the acculturation of EA Amer-

icans but also indicates that acculturation may not be sufﬁciently

strong. Although EA Americans are native English speakers raised

in the United States, their cultural upbringings can still predispose

them to high ethnic homophily and thus disadvantage them in

leadership emergence.

Ethnic Homophily Negatively Predicts Leader

Nominations for Each Ethnic Group

We also tested whether ethnic homophily negatively predicted

leadership emergence for each of the four ethnicities. To maximize

statistical power, we combined samples from Studies 2 and 3.

In further support of Hypothesis 1, ethnic homophily negatively

predicted leader nominations for each of the four ethnicities (with

controls): EAs (B=−3.36, SE =.83, p<.001), SAs (B=−3.24,

SE =.75, p<.001), Latinos (B=−5.10, SE =.66, p<.001), and

Whites (B=−.57, SE =.21, p=.008). That is, among EAs, the

more ethnically heterophilous EAs were more likely to emerge as

leaders. Even among Whites (ethnic majority), the more ethnically

heterophilous Whites were more likely to emerge as leaders.

Together, these results further demonstrate that, regardless of

one’s ethnicity, bonding with individuals from other ethnic back-

grounds is conducive to leadership emergence in multiethnic

environments.

General Discussion

The three studies supported our ethnic homophily perspective on

the Bamboo Ceiling phenomenon. Analyzing a large survey of

54,620 JD students from 124 U.S. law schools, Study 1 found that

EAs displayed higher ethnic homophily than all other ethnicities

(SAs, Whites, Blacks, Latinos, Middle Easterners, and Native

Americans). Using social network analysis, Studies 2 and 3 again

found that, compared to SAs, Latinos, and Whites, EAs exhibited

higher ethnic homophily, which mediated EAs’lower leadership

emergence in a U.S. business school. Consistent with recent

research (Lu et al., 2020), EAs were also lower on assertiveness,

which independently mediated their low leadership emergence.

Simultaneous mediation analyses found that high ethnic homo-

phily and low assertiveness both accounted for EAs’disadvan-

tage in leadership emergence. These ﬁndings were reliable (a) for

both EA internationals and EA Americans, (b) across both

survey and network measures of ethnic homophily, (c) whether

we assessed leadership emergence with leader nominations or

leader election, and (d) after we accounted for a host of control

variables (e.g., network centrality, English proﬁciency, person-

ality, demographics).

Theoretical Contributions

The present research offers three major theoretical contributions.

First, we contribute to cultural psychology by advancing the nascent

literature on the Bamboo Ceiling phenomenon. Whereas a wealth of

research has examined the Glass Ceiling phenomenon for women

(Morrison et al., 1987), limited research has examined the Bamboo

Ceiling phenomenon. Our research extends this literature by pro-

viding insights into both the scope and mechanisms of the Bamboo

Ceiling phenomenon. In terms of scope, we found that ethnic EAs—

but not ethnic SAs—were less likely than other ethnicities to emerge

as leaders in multiethnic environments. This ﬁnding suggests that

the Bamboo Ceiling is not an issue for all Asians, but a cultural issue

unique to EAs. By culturally distinguishing between EAs and SAs,

the current research moves beyond the predominant West-vs-East

paradigm (Gelfand & Denison, 2020;San Martin et al., 2018)

and highlights the importance of examining differences within

the Asian ethnic umbrella. Future research should theoretically

and empirically distinguish among Asian subgroups where possible

(e.g., when collecting demographic information and conducting

analyses). In terms of mechanisms, we revealed a theoretically

novel explanation for the Bamboo Ceiling phenomenon. Whereas

past research has focused on EAs’low assertiveness as a mechanism

for their disadvantage in leadership emergence (Lu et al., 2020), we

uncovered ethnic homophily as another mechanism above and

beyond assertiveness. That is, EAs are less likely to emerge as

This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Figure 5

Study 3: Simultaneous Mediation Analysis for Leader Election

18 LU

Template Version: 11 October 2021 ▪6:01 pm IST PSP-A-2020-1216_format_ﬁnal ▪8 November 2021 ▪3:53 pm IST

leaders partly because their ethnic homophily is culturally incon-

gruent with the ethnically heterophilous leader suited to multiethnic

environments. Importantly, the Bamboo Ceiling occurred for both

EA internationals and EA Americans: Both were more homophilous

than their non-EA counterparts and thus less likely to emerge as

leaders in multiethnic environments.

Second, we contribute to the leadership literature by advancing a

social network perspective on leadership emergence in multiethnic

environments. Whereas much research has examined how social

network inﬂuences leadership effectiveness, less research has exam-

ined how social network inﬂuences leadership emergence (Lu,

2018), especially in multiethnic environments (see Carter et al.,

2015, for a review). Our network studies consistently found a

negative link between ethnic homophily and leadership emergence

for each of the ethnicities—that is, not only for EAs, but also for

SAs, Latinos, and Whites. Regardless of ethnicity, ethnically het-

erophilous individuals were more likely to emerge as leaders in

multiethnic environments. These results suggest that bonding with

people from different ethnic backgrounds is conducive to leadership

emergence in multiethnic environments, which are increasingly

common in the contemporary society.

Third, we contribute to the literature on ethnic homophily.

Whereas past studies have mostly compared the ethnic homophily

of Whites versus non-Whites (Mollica et al., 2003), we provided a

more comprehensive and nuanced comparison among different

ethnic groups. By contrasting the ethnic homophily of EAs and

other ethnicities, we highlight the role of culture in shaping social

network dynamics. Drawing on cultural differences in communica-

tion style, ethnic heterogeneity, and relational mobility, we theo-

rized why EAs may be more ethnically homophilous than other

ethnicities. We not only revealed cultural differences in ethnic

homophily but also demonstrated how they translate into differences

in leadership attainment—an important consequence. In doing so,

we extend the literature on the double-edged nature of ethnic

homophily: While same-ethnicity ties can provide social support

(Ibarra, 1993;James, 2000), they can also impede individuals’

leadership emergence in multiethnic environments. By examining

the sociological construct of homophily, we respond to the call to

integrate social network analysis into psychology (Burt et al., 2013;

Kilduff & Lee, 2020;Kitayama, 2017;Oishi et al., 2009).

Practical Implications

The present research has actionable practical implications for

society, organizations, and individuals in an increasingly multieth-

nic world. Society should be alert to EAs’underrepresentation in

leadership roles—rather than assuming that Asians are the model

minority “doing just ﬁne”(Lu et al., 2020). To help EAs break

through the Bamboo Ceiling, organizations could create opportu-

nities for interethnic interactions, such as interethnic lunches or

mentorships. Indeed, research suggests that social contact is a

prerequisite for meaningful relationships (Festinger et al., 1950;

Reagans, 2011). For EA individuals themselves, it is important to be

cognizant that high ethnic homophily can contribute to their under-

representation in leadership roles. EAs may beneﬁt from a conscious

effort to diversify their social networks (e.g., sports with individuals

from other ethnic backgrounds).

More broadly, our research provides insights into an important

question: How can individuals emerge as leaders in multiethnic

environments? Consistent with the literature (McPherson et al.,

2001), we found that all ethnicities exhibited ethnic homophily

to some extent. To emerge as leaders in multiethnic environments,

individuals need to step out of their cultural comfort zone and bond

with others from different ethnic backgrounds.

Limitations and Future Directions

The current research has several limitations which provide op-

portunities for future research. First, although Study 3 leveraged a

time-lagged design to examine the effect of ethnic homophily on

objective leadership emergence 1 year later, strict causal inference

could not be established. Future research could examine whether

experimental interventions that help EAs develop friendships with

ethnic outgroups can increase their leadership emergence (Gino

et al., 2020). For example, acculturation research suggests that

individuals who integrate the host culture into their ethnic identity

tend to exhibit lower ethnic homophily in friendship networks

(Mok et al., 2007;Repke & Benet-Martínez, 2017).

Second, although our law school study surveyed a large sample of

JD students and our MBA studies enabled us to examine the natural

processes of friendship formation and leadership emergence in a

new multiethnic environment (i.e., high ecological validity), future

research should ascertain the generalizability of our ﬁndings in other

settings. For example, researchers could examine whether EA

employees are higher on ethnic homophily than non-EA employees,

and are thus less likely to emerge as leaders in the workplace.

Third, while our theory section discussed cultural differences in

(a) communication style, (b) ethnic heterogeneity, and (c) relational

mobility as potential explanations for EAs’high ethnic homophily,

our studies did not test these explanations. Additionally, while EAs

may prefer to socialize with other EAs, non-EAs’biases can also

contribute to EAs’ethnic homophily (Lin et al., 2005;Mehra et al.,

1998).

Future research is needed to pinpoint what mediates the link

between our predictor (EA vs. non-EA) and mediator (ethnic

homophily). Similarly, future research is needed to pinpoint what

mediates the negative link between our mediator (ethnic homophily)

and outcome (leadership emergence).

Fourth, it is noteworthy that our study contexts are multiethnic

environments, where socializing with people from different ethnic

backgrounds is conducive to leadership emergence. We speculate

that ethnic homophily is less likely to impede leadership emergence

in mostly monoethnic environments, as monoethnic environments

do not require a leader to bond with ethnic outgroups. For example,

in monoethnic cultures like Japan, it is essential for a Japanese

individual to bond with other Japanese in order to attain leadership.

This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Notably, Lu et al. (2020) tested prejudice as a potential mechanism for

EAs’underrepresentation in leadership. They operationalized prejudice as

“affective social distance, or antipathy toward close interactions with

members of a group”(p. 4597). The researchers measured prejudice with

items such as “How comfortable would you be if a [Chinese/Japanese/

Korean/Bangladeshi/Indian/Pakistani] American dated your sibling/shared

an ofﬁce cubicle with you/became your next-door neighbor ::: ?”(1 =very

uncomfortable,6=very comfortable). Lu et al. (2020) consistently found

that EAs faced less prejudice than SAs on average, “possibly because of SAs’

darker skin tone and physical resemblance to certain Middle Easterners.

For instance, SAs experienced considerable ethnic hostility in the aftermath

of 9/11”(p. 4591). If EAs similarly faced less prejudice than SAs in our

studies, then anti-EA prejudice would be unlikely to be the main driver of

EAs’high ethnic homophily.

A SOCIAL NETWORK PERSPECTIVE ON THE BAMBOO CEILING 19

Template Version: 11 October 2021 ▪6:01 pm IST PSP-A-2020-1216_format_ﬁnal ▪8 November 2021 ▪3:53 pm IST

Finally, it is noteworthy that the multiethnic contexts in our

studies are nonpolitical. While ethnic homophily may negatively

predict individuals’leadership emergence in non-political contexts,

ethnic homophily may produce cohesiveness that enables bloc

voting for ethnic minority candidates in political contexts

(Wolﬁnger, 1965).

Conclusion

The present research advanced a social network explanation for

the Bamboo Ceiling phenomenon: ethnic homophily. We analyzed

a large-scale survey of JD students from 124 U.S. law schools and

conducted social network analysis on 11 complete class sections of

new MBA students in a U.S. business school. Our studies revealed

that ethnic East Asians—but not ethnic South Asians—were less

likely than other ethnicities to emerge as leaders, partly because

EAs socialize more with ethnic ingroup members (i.e., other EAs).

The Bamboo Ceiling phenomenon arises partly from the cultural

mismatch between EAs’high ethnic homophily and the ethnic

heterophily needed for leadership emergence in multiethnic

environments.

References

Adair, W. L., Buchan, N. R., Chen, X.-P., & Liu, D. (2016). A model of

communication context and measure of context dependence. Academy of

Management Discoveries,2(2), 198–217. https://doi.org/10.5465/amd

.2014.0018

Adler, N. E., Epel, E. S., Castellazzo, G., & Ickovics, J. R. (2000).

Relationship of subjective and objective social status with psychological

and physiological functioning: Preliminary data in healthy white women.

Health Psychology,19(6), 586–592. https://doi.org/10.1037/0278-6133

.19.6.586

Alesina, A., Devleeschauwer, A., Easterly, W., Kurlat, S., & Wacziarg, R.

(2003). Fractionalization. Journal of Economic Growth,8(2), 155–194.

https://doi.org/10.1023/A:1024471506938

Bai, X., Ramos, M. R., & Fiske, S. T. (2020). As diversity increases, people

paradoxically perceive social groups as more similar. Proceedings of the

National Academy of Sciences of the United States of America,117(23),

12741–12749. https://doi.org/10.1073/pnas.2000333117

Balkundi, P., & Kilduff, M. (2005). The ties that lead: A social network

approach to leadership. The Leadership Quarterly,16(6), 941–961.

https://doi.org/10.1016/j.leaqua.2005.09.004

Ballinger, G. A., Cross, R., & Holtom, B. C. (2016). The right friends in the

right places: Understanding network structure as a predictor of voluntary

turnover. Journal of Applied Psychology,101(4), 535–548. https://

doi.org/10.1037/apl0000061

Bendersky, C., & Shah, N. P. (2013). The downfall of extraverts and rise of

neurotics: The dynamic process of status allocation in task groups.

Academy of Management Journal,56(2), 387–406. https://doi.org/10

.5465/amj.2011.0316

Berdahl, J. L., & Min, J.-A. (2012). Prescriptive stereotypes and workplace

consequences for East Asians in North America. Cultural Diversity &

Ethnic Minority Psychology,18(2), 141–152. https://doi.org/10.1037/

a0027692

Bloomberg. (2017). Asian Americans climb the corporate ladder, but only so

high.https://www.bloomberg.com/news/articles/2017-11-21/asian-americans-

climb-the-corporate-ladder-but-only-so-high

Bonacich, P. (2007). Some unique properties of eigenvector centrality.

Social Networks,29(4), 555–564. https://doi.org/10.1016/j.socnet.2007

.04.002

Burt, R. S., Kilduff, M., & Tasselli, S. (2013). Social network analysis:

Foundations and frontiers on advantage. Annual Review of Psychology,

64(1), 527–547. https://doi.org/10.1146/annurev-psych-113011-143828

Carter, D. R., DeChurch, L. A., Braun, M. T., & Contractor, N. S. (2015).

Social network approaches to leadership: An integrative conceptual

review. Journal of Applied Psychology,100(3), 597–622. https://

doi.org/10.1037/a0038922

Cheryan, S., & Monin, B. (2005). “Where are you really from?”: Asian

Americans and identity denial. Journal of Personality and Social Psy-

chology,89(5), 717–730. https://doi.org/10.1037/0022-3514.89.5.717

Chin, M. M. (2020). Stuck: Why Asian Americans don’t reach the top of the

corporate ladder. NYU Press. https://doi.org/10.1093/sf/soaa117

Chung, E., Dong, S., Hu, X. A., Kwon, C., & Liu, G. (2017). A Portrait of

Asian Americans in the Law. Yale Law School and National Asian Paciﬁc

American Bar Association; https://www.apaportraitproject.org

De, S., Gelfand, M. J., Nau, D., & Roos, P. (2015). The inevitability

of ethnocentrism revisited: Ethnocentrism diminishes as mobility

increases. Scientiﬁc Reports,5(1), Article 17963. https://doi.org/10

.1038/srep17963

DeYoung, C. G. (2006). Higher-order factors of the Big Five in a multi-

informant sample. Journal of Personality and Social Psychology,91(6),

1138–1151. https://doi.org/10.1037/0022-3514.91.6.1138

Digman, J. M. (1997). Higher-order factors of the Big Five. Journal of

Personality and Social Psychology,73(6), 1246–1256. https://doi.org/10

.1037/0022-3514.73.6.1246

Edwards, J. R. (2008). Person–environment ﬁt in organizations: An assess-

ment of theoretical progress. The Academy of Management Annals,2(1),

167–230. https://doi.org/10.5465/19416520802211503

Ertug, G., Gargiulo, M., Galunic, C., & Zou, T. (2018). Homophily and

individual performance. Organization Science,29(5), 912–930. https://

doi.org/10.1287/orsc.2018.1208

Falk, C. F., Heine, S. J., Yuki, M., & Takemura, K. (2009). Why do

Westerners self-enhance more than East Asians? European Journal of

Personality,23(3), 183–203. https://doi.org/10.1002/per.715

Fearon, J. D. (2003). Ethnic and cultural diversity by country. Journal of

Economic Growth,8(2), 195–222. https://doi.org/10.1023/A:10244

19522867

Feiler, D. C., & Kleinbaum, A. M. (2015). Popularity, similarity, and the

network extraversion bias. Psychological Science,26(5), 593–603. https://

doi.org/10.1177/0956797615569580

Fernandez, R. M., Castilla, E. J., & Moore, P. (2000). Social capital at work:

Networks and employment at a phone center. American Journal of

Sociology,105(5), 1288–1356. https://doi.org/10.1086/210432

Festinger, L., Schachter, S., & Back, K. (1950). Social pressures in informal

groups: A study of human factors in housing. Harper.

Flynn, F. J., Reagans, R. E., & Guillory, L. (2010). Do you two know each

other? Transitivity, homophily, and the need for (network) closure.

Journal of Personality and Social Psychology,99(5), 855–869. https://

doi.org/10.1037/a0020961

Frey,L.L.,&Roysircar,G.(2006).SouthAsianandEastAsianinterna-

tional students’perceived prejudice, acculturation, and frequency of

help resource utilization. Journal of Multicultural Counseling and

Development,34(4), 208–222. https://doi.org/10.1002/j.2161-1912

.2006.tb00040.x

Friedman, R., Kane, M., & Cornﬁeld, D. B. (1998). Social support and career

optimism: Examining the effectiveness of network groups among black

managers. Human Relations,51(9), 1155–1177. https://doi.org/10.1177/

001872679805100903

Gaol, F. L., Matsuo, T., & Maulana, A. (2019). Network model for online

news media landscape in Twitter. Information (Basel),10(9), Article 277.

https://doi.org/10.3390/info10090277

Gee, B., & Peck, D. (2018). Asian Americans are the least likely group in

the U.S. to be promoted to management. Harvard Business Review.

This document is copyrighted by the American Psychological Association or one of its allied publishers.

A Social Network Perspective on the Bamboo Ceiling: Ethnic Homophily Explains Why East Asians but Not South Asians Are Underrepresented in Leadership in Multiethnic Environments

Abstract and Figures