Conversion among Chinese Overseas Students in the US: A Choice Model on Individual Characteristics and Organizational Traits

Abstract

After people migrate to a different religious landscape, they will have multiple candidate religions to choose from, and some people may convert to a new religion. This study argues that both individual-level characteristics and a religion’s organizational traits are involved in religious conversion. Using cross-sectional samples of Chinese overseas students and scholars in the US in 2018 (n = 1911), we deployed mixed multinomial models to demonstrate how conversion, measured as intergenerational and personal religious changes, is associated with individual-level factors and organizational traits of religions. We found that the choice between Christianity, Buddhism, and Chinese folk religion, compared with no religion, is associated with unique individual-level characteristics. For religious organizational traits, missionary intensity and organized activity intensity are generally associated with a higher likelihood of conversion. This study distinguishes the different levels of the operating mechanism in conversion and points out an interactive and heterogenous model for individuals’ choice of various competing religions.

1. Introduction

Religion is a system that designates the division of the sacred relative to the profane and regulates practices among a group of individuals holding solidarity and forming a moral community (Durkheim [1912] 2008). Religious conversion refers to the changes in religious identity and belonging. It may happen intergenerationally, when religious adherence of an offspring differs from that of their parents, or occur within the life course of an individual (Breen and Hayes 1996; Hu and Leamaster 2015; Sherkat and Wilson 1995). Theoretically, voluntary religious change requires a religious organization to have certain characteristics and traits that an individual is attracted to. Beyond individual characteristics, scholars have pointed out (Gooren 2007; Hall 2006; Stark and Finke 2000; Wang and Yang 2006; Yang 2010a) that the characteristics of a religious organization, such as its organized provision of religious resources, culturally compatible messages, religious service quality, and intensity, make up the fundamental attraction for potential converters. Hence, they have called for determining the factors behind conversion at three distinct levels: individual, organizational, and contextual (Yang and Abel 2014).

In other words, people with certain characteristics make a choice on the basis of the utility of that choice relative to other options. Because utility is usually latent and unobservable, scholars may model it as a function of the organizational traits of that particular religion. For example, Sherkat (2014, 1995) emphasized the organizational and socioeconomic arrangement behind conversion in the US. In summary, to assess conversion, both individual impetus and the favorable characteristics of a religious denomination should be considered in tandem.

In a separate line, the literature on the conversion process also argues that a threshold must be overcome to reach a religious identity. Without a strong pulling force by an active religious organization, individuals with suitable conditions for conversion (e.g., weak ties to the origin religion) only vacillate around a liminal stage, without converting. Lim et al. (2010) noticed a substantial number of “liminal religious nones” in the US vacillate between religious adherence and no religion. In an earlier study, Lofland and Stark (1965) elaborated in their process model that conversion involves renouncing an old faith and adopting a new one. These studies on the conversion process implied that two types of conditions must both be met for individuals to transition to a different faith: some individual contingencies such as personality, experiences, and ties to family predispose certain people and put them at a greater risk of renouncing an old faith; meanwhile, the candidate religion must also reach a person’s level of sufficient attraction to successfully recruit them, which can be proxied through missionary works, outreach, cultural compatibility, and other organizational factors.

Thus, we conceptualize conversion according to the two-sides model: an individual’s background characteristics (e.g., socioeconomic status, family connection, prior beliefs, and sociopolitical attitudes) create a propensity condition in favor of choosing a specific candidate religion. Meanwhile, the individual’s background characteristics need to be combined with the organizational traits of the candidate religions, such as missionary works and organized activities. To demonstrate the applicability of this model, this study uses a sample of Chinese overseas students and scholars living in the US and analyzes their choice among belief systems (Christianity, Buddhism, Chinese folk religion, and nonreligion) by interactively modeling individual-level and organization-level factors.

1.1. The Individual Factors in Conversion

We consider in this section the significant roles that family ties, acculturation and discrimination, sociopolitical attitudes, socioeconomic status (SES), and gender play in shaping the individual-level propensity toward conversion among immigrants.

Choosing to adhere to a religious denomination is highly path dependent. The family institution recreates its own religious practices, and individuals tend to inherit their parents’ religious affiliation. An individual’s upbringing community provides a set of religious beliefs and normative worldviews specific to a religious tradition, rendering a person more likely to attend more religious services with them and adopt, to varying degrees, beliefs from significant social others (Gooren 2007; Hall 2006; Snow and Machalek 1984). On the other hand, people with weaker attachments to their original religious or cultural community display a greater tendency toward choosing a new belief system. The disruption of home networks is a scenario that appears particularly relevant for many migrants, some of whom will eventually embrace more-innovative cultural elements from the host environment and ultimately convert to a new religion as they become more acculturated (Berry 1997; Cadge and Howard Ecklund 2007; Chen 2005). In empirical research, studies on religious choice and mobility need to control for factors such as one’s previous religious identification, the religious identification of one’s parents, and the strength of familial ties. In short, we hypothesize that stronger family ties are inversely associated with the likelihood of conversion among this body of international personnel.

Socioeconomic status (SES) constitutes an important layer of individual background for potential conversion, although many have found the association between SES and religiosity tends to be bidirectional. The opium thesis of religion contends that people with lower SES find greater psychological compensation in religious ideas. The opium thesis originates from Marx’s claim that “religion is the opium of the masses”. Schnabel (2020) showed that those lacking social status and economic prosperity use religion as a compensatory resource. Whether by providing compensatory resources, creating a normative status, or generating more ties and social capital, religion constitutes a social focus with an abundance of resources and information embedded in its membership pool (Becker and Dhingra 2001; Maselko et al. 2011; Putnam 2001; Wuthnow 2002). This feature may attract individuals who want a particular type of asset to the provider of this asset (e.g., a religious organization), and they may convert to a different religion for socioeconomic improvement. Meanwhile, the relationship between SES and religiosity is far from definitive, as people with a better SES have more-frequent religious attendance and a higher prevalence of religious identification, at least in contexts where religions have attained social desirability (Goode 1966; Lazerwitz 1961; Wuthnow 2002). In regard to SES, we will test the hypothesis formulated by the opium thesis: SES is inversely associated with the likelihood of choosing a different religion relative to one’s original religion.

People endowed with certain ascribed identities, such as gender and race, may disincline to fixate on a religion. Lévi-Strauss (1969) and modern applicants of the structural theory (Doja 2008) explain how women are symbolically and socially relegated to a marginal position in a patriarchal cultural system, leading to their weaker conviction in a particular culture and be more receptive to innovative worldviews. Thus, female immigrants are more likely to negotiate for a renewed social status by converting to a new religious/cultural system (Carnes and Yang 2004; Chen 2005; Marquardt 2005). Racial and ethnic minorities under oppression sometimes also escape from a region’s dominant religion to form their distinct identity and communal independence (Curtis 2012; Gokhale 1986). Thus, we propose that being a woman or an ethnic minority increases the likelihood of choosing a different religion relative to one’s original religion.

Another series of important factors in conversion relates to an individual’s sociopolitical attitudes. Such individual-specific attitudinal and psychological characteristics include experiences of racism and discrimination (Bowen 2013; Curtis 2012; Gokhale 1986; Robinson and Clarke 2003), political affiliation, and liberalism (Audette et al. 2017; Davis and Perry 2021; Whitehead and Perry 2020). In short, one’s sociopolitical attitudes, such as the authoritarian–liberal spectrum, racial attitude, and gender views, may influence the likelihood of conversion at the individual level. We will treat these attitudinal variations as control variables that may confound one’s choice process.

1.2. The Organizational Factors in Conversion

The rational actor theory of religion expects that a religious organization with greater intensity in organizing and disseminating religious resources becomes more productive in recruitment (Sherkat and Wilson 1995; Stark and Finke 2000; Stark and Iannaccone 1994). The organizational strength of a religion comprises its financial resources, and it builds on a repertoire of cultural resources and a mobilization capacity for recruiting converts and managing its congregation (Sherkat and Wilson 1995). Some religious organizations may simply be more proactive at engaging in proselytism, outreach, and different religious service provision. For example, a larger number of lapsed and dropout Catholics in some countries were accounted for by the idea of religious monopoly, which historically led the Church to be less enthusiastic about proselytism and outreach (Stark and Finke 2000). The intensity of outgoing religious activities, particularly interpersonal proselytism, also explained the decline of mainline Protestantism and the relative rise of evangelical Protestants starting in the mid 20th century (Stoll and Petersen 2008). The rational actor theory explains the pattern of religious switching particularly well in democratic contexts where conversion incurs little cost and stigma (Barro et al. 2010). This theoretical framework corroborates our argument that a religion’s organizational traits manifested through factors such as missionary works, member enthusiasm, and activity frequency may increase the chance that a pondering individual is pulled toward the religion.

Another important discourse on the organizational foundation of conversion relates to how a religion’s growth benefits from its compatibility with the prevailing cultural environment (Glock and Bellah 1976; Wuthnow 1976) and thus gains higher collective salience. A more salient religion also has more members willing to disclose their identity. Such a religion with higher collective salience may exert greater exposure opportunities for potential seekers to learn about its messages. People from religions with greater collective cultural salience may also have less stigma to evangelize and recruit new members. The collective salience of religion has been shown to attract new members in various cultural contexts. The state religion of a colonial state emits favorable symbolic value and enjoys greater salience among its colonial subjects (Segovia 2000). Goossaert and Palmer (2011) analyzed how the dimming of traditional Chinese religions was a consequence of renewing political and social institutions, rather than seeing the fluctuation of religious growth as a deviation from an ideal norm. When disentangling the puzzle that former communist countries have undergone explosive growth in religions and religious experiences, Yang (2005, 2010a) emphasized the perceived salience of Christianity as a religion better suited for modernization. Overall, under the condition that an individual with specific characteristics has reached the threshold for conversion, a religion with more favorable organizational traits relative to its competitors is seen as yielding greater utility and is this more likely to be chosen. We propose that given the same set of individual characteristics, the organizational traits of a religion, including missionary intensity, activity intensity, and collective salience, are associated with a greater likelihood of people’s choosing this religion.

1.3. Gap in the Literature

There are two main issues underlying the gap in the current literature on conversion. The first issue concerns the need for both individual and organizational factors in completing the full process of conversion. The second issue concerns the negligence that the choice in conversion can usually be made on multiple candidate religions, not a single destination.

First, as reviewed, both individual background characteristics and a religion’s organizational traits substantially affect the choice of a pondering individual. Second, the organizational traits of a religion may also attract people with certain endowments, necessitating scholars to consider the interaction between individual and institutional factors. For example, Weber and Merton both discussed the elective affinity between Protestant ethics and an entrepreneurial devotion in sciences and business, laying out the interpretive pathway that certain characteristics of Calvinism attract people with such a predisposition (Merton 1936; Weber [1905] 2002).

However, most quantitative studies on conversion still focus on individual psychosocial elements, while most studies with an organizational focus fall into various types of case studies and comparative-historical analyses. Quantitative empirical studies on conversion are few. Informed of the significance of the synergetic forces of individual and organizational factors, we believe that an analysis of conversion should consider the mechanisms at both ends of the conversion process and extend beyond individual psychosocial elements.

The second issue concerns the potential choice between candidate religions. Potential converters have a choice among different religions, but extant studies often focus on a single destination religion. The literature on conversion among immigrants to Western countries often fixates on Christianity. This leaves a false impression that religions outside the Judeo-Christian tradition seldom gain members and ignores the fact that individuals are often presented with an array of religions to observe, learn from, participate in, and eventually convert to. On the contrary, the presence of several candidate religions has always been a common experience in China and East Asia. Beyond Christianity, studies have shown that Chinese migrants convert en masse to Theravada Buddhism in Thailand (Carnes and Yang 2004) and to Islam in the Gulf countries (Wang 2020), and some showed sprouting interests in Judaism.1 In the US, immigrants also convert to non-Christian religions such as Buddhism and Islam (Bowen 2013; Leamaster 2012; Yang et al. 2018). In addition, another seldom-attended but important population comprises those moving to nonreligions (Cavalcanti and Schleef 2005). If we regard different religions as candidates in an election where the choice of one necessarily means the discard of another, then the remaining question becomes what accounts for a person’s choice of one religion over other available options.

The current study intends to resolve the above two issues by treating all major religious traditions, namely Chinese folk religions, Buddhism, Christianity, and nonreligion, as competing candidates among the Chinese overseas students. Unique to this conceptualization is our assumption that the choice of a religion is a multinomial evaluation of the relative attractiveness against all available candidate religions. First, we evaluate the likelihood of choosing a religion on the basis of its organizational traits relative to other candidate choices. These organizational traits are organized activity intensity, missionary intensity, and collective salience. Second, we estimate how individual-level variables are associated with the likelihood of choosing a religion relative to other candidates and whether the individual-level effects are heterogeneous for each religion.

2. Methodology

2.1. Sample

The data set comes from cross-sectional surveys conducted in 2016 and in 2018 among Chinese overseas students at two universities in the US. One university is located in the US Midwest and the other in the US Deep South. Owing to the disproportionally smaller size of Chinese students at the southern university (3.2% of the sample), samples from two universities were combined without harming the heterogeneity. Recruitment was conducted online and initiated by the emails sent to the entire target population from authorities in charge of international student affairs at each university, which voids the concern about sampling error. Three reminders sent to the population during the next 2 weeks also reduced nonresponse bias. The overall response rate was 22.4%, full completion rate was 77%, and we had considerably overachieved our initially needed sample size by having 916 respondents in the 2016 survey and then 767 respondents in the 2018 survey. Our sample is also closely representative of the population regarding the marginal totals by gender, college, and seniority. More details on the survey can be found in serial reports at www.globaleast.org/publications/survey-reports/ (accessed on 7 April 2022).

2.2. Mixed Multinomial Model

As argued in the introduction, migrants facing the potential option of conversion have more than one option to choose from. The conventional analysis targets a single religion and thus overlooks the alternatives relative to each decision. Therefore, we combined conditional logit regression, alternatively known as the discrete choice model, with multinomial logit regression to fit our data to the conceptual framework. First applied to consumer choice behavior and then widely to other social behaviors, the conditional logit regression considers the characteristics of each choice object and estimates the likelihood of each choice relative to other choices (Hoffman and Duncan 1988; McFadden 1974). Consider that person i has a number of choices to make: {a, b, … z}. If the utility function of choice a is the highest among all alternatives, u(a) > ∑u(−a ∈ {a…z}), then we need to model which option yields the best utility. Because the utility of a choice is an unobservable “implicit knowledge”, one can model it only as a likelihood function of an explicit choice among a list of alternatives given several covariates. Regression models with a single outcome as the function of the covariates miss this essential feature of choice behaviors; that is, people can always switch to alternatives.

The conditional logit model and multinomial logit model belong to the same general class, the main distinction being that the conditional logit model has a constant coefficient parameter for all candidates of choice, whereas in the multinomial logit model, such a parameter is individual specific. The probability that a person will choose dth religion over J candidates, conditional on dth religion’s characteristics x, is:

$$Pr\left(d|x_d\beta \right)=\frac{\exp \left(x_d\right)+{ϵ}_d}{{\sum }_i^j\exp \left(x_i\beta \right)+{ϵ}_i}, \forall {ϵ}_d, {ϵ}_i~\left(0,\sigma \right)$$

Beyond the characteristics of each candidate choice, when we add information about individual respondents, the conditional logit regression is combined with multinomial logit regression to become a mixed multinomial model. The mixed multinomial model extends the conditional logit regression from the above equation by multiplying the summated likelihood of choosing dth religion with an individual-level variable $u_k$ for the kth individual:

$$Pr\left(d_k|x_d\beta +{\eta }_k\right)=\frac{\exp \left(x_d\beta \right)+{ϵ}_d}{{\sum }_i^j\exp \left(x_i\beta \right)+{ϵ}_i}\ast \exp \left(u_k\gamma \right)+{\eta }_k, \forall {\eta }_k,{ϵ}_d, {ϵ}_i~\left(0,\sigma \right)$$

As $u_k$ describes a set of covariates at the individual level, we can consider the probability of choosing dth religion in a mixed multinomial model as a conditional probability that has incorporated two types of information: (1) the probability is relative to all other J candidate religions and (2) the probability is conditional on individual-level covariates. When these individual-level covariates include prior religious affiliations, the choice of dth religion is an outcome net of, or minus, one’s prior religious affiliation in the dth religion, so the effect expresses conversion.

2.3. Measurement

In the choice of a religious denomination ($d_k$), to fit the data with the above model, we estimate the parameters by measurements from the survey. The outcome variable $d_k$ refers to a list of religious traditions to which the respondents indicated they belonged at the survey time. Because of the ambiguity in the classification of religions in the Chinese sociopolitical context and the substantial number of religious nones (Hu and Leamaster 2015; Yang 2006), we did not explicitly elicit answers from a list of religious categories but instead asked respondents to rate their level of belief in different religious denominations. This novel measure of religious identification has been applied to accommodate the syncretic nature of the Chinese religious landscape (Hu and Leamaster 2015; Yang and Yang 2017). Subsequently, five mutually exclusive religious identities were constructed on the basis of the rating of those six religious denominations, as illustrated in

Table 1. Construction of religious identification.

Choosing “More Believing” or “Entirely Believe” in Either:	Buddhist	Folk Syncretic2	Christian	Religious None
Buddhism	Yes	-	No	No
Taoism	No	Yes	No	No
Folk religions	No	Yes	No	No
Islam3	No	-	No	No
Protestantism	No	-	Yes	No
Catholicism	No	-	Yes	No

: Buddhists are those who “believe more” or “believe entirely” only in Buddhism; Christians refer to the respondents who “believe more” or “believe entirely” only in Protestantism or Catholicism; Muslims comprise the sampled participants who “believe more” or “believe entirely” only in Islam; Chinese folk religionists are those who “believe more” or “believe entirely” in Taoism, Confucianism, or folk religions. This is because Chinese folk religion is a system that flexibly adapts to changing environments (Weber and Gerth 1953) and embraces a syncretic treatment of the elements of different religious traditions (Yang and Hu 2012). Religious nones are those who do not believe in any religious denominations.

Conversion is a change in identification from a previous one. In the literature, both personal conversion and intergenerational conversion are important topics, with ethnographic studies tilting toward personal religious conversion experience, whereas many quantitative demographic studies focus on intergenerational religious change. China’s rapid religious change is marked by intergenerational conversion, where the younger generation is more religious than the older ones born under Maoism (McPhail and Yang 2020; Yang 2005, 2010b). In addition, the religious change in the US is also marked by generational differences: younger cohorts are less religious, and the religious socialization of children has become less effective thanks to interreligious marriage and other factors (Sherkat 2014; Voas and Chaves 2016).

Therefore, this study analyzes two types of conversion: personal and intergenerational. Personal conversion refers to a choice of a religion relative to a person’s premigration religious identification, while intergenerational conversion refers to an individual’s choice of religion relative to their parent’s religious identification. In addition to an individual’s current religious identity, the survey asked about respondents’ belief ratings on six denominations as they recalled prior to their coming to the US. In Table 1, five religious identities are categorized according to these principles.

For intergenerational mobility, we contrasted the survey question “What was the religious identity of your father when you were 15?” with the survey question “What was the religious identity of your mother when you were 15?”, where the options were “Buddhism, Taoism, Protestantism, Catholicism, Islam, folk religions, others, no religion”. The age of 15 was chosen according to the prevailing evidence that parental religious influence stabilizes during adolescence, between ages 15 and 16 (Francis 1993; Martin et al. 2003). We categorized parental religious identity into the same five types as the child’s religious identity: Buddhism, Chinese folk religions, Christianity, Islam, and nonreligion. Whether it is premigration religious identity, paternal religious identity, or maternal religious identity, we can estimate conversion when each is treated as a fixed-effect categorical variable.

Table 2. Cross-tabulated comparison of religious conversion, by intergenerational and personal changes in religious identities.

	Current Religion
Paternal Religion, n = 1512	Buddhist	Folk Religionist	Christian	Religious None
Buddhist	38	40	8	50
Folk syncretic	2	14	7	23
Christian	1	4	15	12
None	87	171	101	885
Maternal religion, n = 1512
Buddhist	52	57	14	79
Folk syncretic	5	10	5	25
Christian	0	8	22	16
None	71	156	90	850
Premigration religion, n = 1911
Buddhist	104	7	10	22
Folk syncretic	12	531	8	20
Christian	1	2	51	7
None	8	10	53	991

shows intergenerational conversion and personal conversion.

Religion-specific characteristics are specific to each religion, not to each respondent, and thus have a degree of freedom equal to the number of candidate religions minus 1. We calculated three indicators (missionary activity, salience, and organized activity) for four religious traditions (Christianity, Buddhism, folk religions, and no religion) by aggregating individual responses on a common survey item that touches base with all respondents’ shared experiences and knowledge. Such measure effectively assesses the property of a lived collective unit (a religious denomination in our case) and has been recognized in research as an indicator of compositional effect (Duncan et al. 1998). Characteristics that pertain to each religion were measured by the following operations: Missionary intensity was assessed by aggregating the respective responses to “had a member from this religion advocated for their religion to you since coming to the US: Buddhism, Taoism, Protestantism, Catholicism, Islam, Mormon, Falun Gong, folk religions?” The organized activity intensity of each religion was assessed by evaluating the collective frequency of participation in Islamic activity, Buddhism activity, Catholic activity, and Protestant activity for each denominational religion; Confucian activity, Taoism activity, and ancestor worship for folk religions; red tourism for Mao worship; and none for religious nones. The salience level of religion is measured by aggregated responses to the question “What percentage of people who go to the same school with you know your religious beliefs?” While religion-specific characteristics are aggregated measures that describe the compositional effects, such compositional effects are social facts irreducible to individuals and widely used in contextual and neighborhood research to represent the effects of units at a higher level (Macintyre et al. 2002).

For individual-specific characteristics, we proposed that demographic variation between individuals may be associated with the differential likelihood of changing to another religion. We measured demographics by gender (male vs. female), age, years since coming to the US, and ethnicity (Han vs. minority). The opium thesis proposes that people of lower SES will be more attracted to religions, so we measured financial strain to test the opium thesis with a question: “what would you say is your financial burden while living in the US: very heavy burden, somewhat-heavy burden, moderate burden, a little burden, very little burden?” A variant of the opium thesis also contends that the sick and weak tend to seek religions, so we controlled for self-reported health. We also measured familial tie strength with a question: “how often do you contact your families in China: daily, weekly, monthly, or less than monthly?” Six sociopolitical attitudinal attributes, including generalized trust, experience with racism, and authoritarianism, were rated on five-point Likert scales: “I trust most people around me”, “I have been unfairly treated because of my race”, “the efficiency of democracy is too low”, “democracy is not good for maintaining social orders”, “democracy is better than all other political systems in spite of its problems”, and “social order is more important than personal freedom”.

3. Results

Table 3. Descriptive statistics of the sample.

	Mean/%	s.d./n
Religion specific:
Missionary intensity	0.22	0.40
Salience	42.8	9.13
Organized activity intensity	1.29	0.27
Individual specific:
Female	43.3%	770
Age	23.4	4.64
Years in the US	3.50	2.96
Ethnicity (Han)	94.7%	1408
Family contact	1.91	0.64
Financial burden	2.22	1.01
Self-reported health	3.58	0.88
Trust	4.04	1.19
Authoritarianism	3.27	0.89
Experience of racism	2.76	1.41

presents the descriptive statistics of the sample. The average missionary intensity across four religious traditions registers at 0.22. Missionary work is most common from Christianity (0.67), followed by folk religions (0.21) and Buddhism (0.07), and no religion engages in no missionary work (0). The average of collective salience is 42.8, with higher salience for no religion (52.6), followed by folk religion (45.3), Christianity (42.15), and Buddhism (28.4). The mean organized activity frequency is 1.29, with higher activity frequency for Christianity (1.81), followed by Buddhism (1.13), folk religions (1.18), and no religion (1.06). Table 3 also contains descriptive statistics for all individual-level covariates.

Table 4. Mixed logit model of intergenerational religious changes between father’s religion and child’s current religion.

Candidate Religion (ref = None)	Model 1			Model 2
	Buddhism	Christianity	Folk	Buddhism	Christianity	Folk
Religion-specific variables
Missionary intensity	0.39 (0.13) **			0.32 (0.13) *
Organized activity intensity	0.32 (0.19)			0.49 (0.27)
Salience	−0.02 (0.03)			−0.01 (0.03)
Individual-specific variables
Sex				0.69 (0.21) ***	0.3 (0.2)	0.15 (0.16)
Age				−0.05 (0.03) *	0.04 (0.02)	−0.03 (0.02)
Length				0.05 (0.05)	0.13 (0.04) **	0.11 (0.04) **
Ethnicity				0.1 (0.51)	−0.15 (0.46)	−0.29 (0.33)
Financial strain				0.07 (0.1)	0.25 (0.1) *	0.12 (0.08)
Health				0.08 (0.12)	0.06 (0.12)	−0.04 (0.09)
Family tie				−0.07 (0.16)	−0.08 (0.16)	−0.04 (0.13)
Authoritarian				−0.02 (0.14)	0 (0.15)	0.07 (0.11)
Racism				0.15 (0.08)	0.03 (0.08)	0.1 (0.06)
Trust				0 (0.09)	0.03 (0.09)	−0.04 (0.07)
Father’s religion (ref = None)
Buddhist	1.99 (0.24) ***	0.31 (0.4)	1.38(0.23) ***	2.03 (0.26) ***	0.23 (0.42)	1.38 (0.24) ***
Christian	−0.16 (1.05)	2.41 (0.4) ***	0.57(0.59)	−0.16 (1.05)	2.38 (0.42) ***	0.51 (0.59)
Folk	−0.15 (0.75)	0.96 (0.45) *	1.09(0.35) **	−0.01 (0.76)	1.02 (0.47) *	1.1 (0.37) **
McFadden R2	0.05			0.07
N	1463			1365

Standard errors in parentheses after the coefficients. * p < 0.05, ** p < 0.01, *** p < 0.001.

and

Table 5. Mixed logit model of intergenerational religious changes between mother’s religion and child’s current religion.

Candidate Religion (ref = None)	Model 1			Model 2
	Buddhism	Christianity	Folk	Buddhism	Christianity	Folk
Religion-specific variables
Missionary intensity	0.42 (0.13) ***			0.36 (0.13) **
Organized activity intensity	0.3 (0.2)			0.46 (0.27)
Salience	−0.02 (0.03)			−0.02 (0.03)
Individual-specific variables
Sex				0.68 (0.21) **	0.24 (0.21)	0.15 (0.16)
Age				−0.06 (0.03) *	0.04 (0.02)	−0.03 (0.02)
Length				0.04 (0.05)	0.12 (0.04) **	0.11 (0.04) **
Ethnicity				0 (0.51)	−0.13 (0.47)	−0.46 (0.33)
Financial strain				0.04 (0.11)	0.23 (0.1) *	0.12 (0.08)
Health				0.11 (0.12)	0.11 (0.12)	−0.03 (0.09)
Family ties				−0.05 (0.17)	−0.06 (0.16)	−0.02 (0.13)
Authoritarian				−0.01 (0.14)	−0.03 (0.15)	0.08 (0.12)
Racism				0.12 (0.08)	0.04 (0.08)	0.09 (0.06)
Trust				0 (0.09)	0.06 (0.1)	−0.04 (0.07)
Mother’s religion (ref = None)
Buddhist	2.04 (0.22) ***	0.48 (0.31)	1.34 (0.2) ***	2.08 (0.23) ***	0.45 (0.33)	1.35 (0.21) ***
Christian	−16.12 (2760.8)	2.59 (0.35) ***	0.96 (0.44) *	−15.2 (1702.2)	2.48 (0.36) ***	0.65 (0.47)
Folk	0.85 (0.51)	0.6 (0.5)	0.73 (0.39)	1.14 (0.53) *	0.76 (0.52)	0.92 (0.41) *
McFadden R2	0.06			0.09
N	1463			1364

Standard errors in parentheses after the coefficients. * p < 0.05, ** p < 0.01, *** p < 0.001.

present results for intergenerational religious change, where the dependent variable is the child’s religion, where the reference is no religion. Table 4 considers intergenerational religious change between children’s and fathers’ religions, whereas Table 5 presents that between children’s and mother’s religions. Next,

Table 6. Conditional logit model of personal religious changes between premigration religion and current religion.

Candidate Religion (ref = None)	Model 1			Model 2
	Buddhism	Christianity	Folk	Buddhism	Christianity	Folk
Religion-specific variables
Missionary intensity	0.6 (0.22) **			0.47 (0.23) *
Organized activity intensity	−0.23 (0.27)			0.14 (0.37)
Salience	0.15 (0.04) ***			−0.01 (0.06)
Individual-specific variables
Sex				0.17 (0.34)	0.31 (0.26)	−0.14 (0.33)
Age				−0.07 (0.04)	0.05 (0.03)	−0.03 (0.04)
Length				0.1 (0.08)	0.16 (0.05) **	0.19 (0.08) *
Ethnicity				−0.08 (0.74)	−0.71 (0.5)	−0.38 (0.69)
Financial strain				0.3 (0.17)	0.26 (0.13) *	0.24 (0.17)
Health				0 (0.2)	0.22 (0.15)	−0.13 (0.19)
Family tie				0.13 (0.26)	0.15 (0.2)	0.04 (0.25)
Authoritarian				0.33 (0.22)	0.09 (0.18)	0.41 (0.24)
Racism				−0.12 (0.14)	−0.01 (0.11)	−0.13 (0.13)
Trust				−0.18 (0.15)	0.03 (0.12)	0 (0.15)
Premigration religion (ref = None)
Buddhist	6.42 (0.43) ***	2.06 (0.41) ***	3.61 (0.55) ***	6.36 (0.45) ***	2.15 (0.43) ***	3.47 (0.63) ***
Christian	2.95 (1.13) **	4.96 (0.43) ***	3.41 (0.87) ***	2.94 (1.15) *	5.24 (0.48) ***	3.6 (0.91) ***
Folk	4.22 (0.52) ***	2.16 (0.45) ***	7.99 (0.4) ***	4.09 (0.54) ***	1.93 (0.48) ***	7.25 (0.46) ***
McFadden R2	0.67			0.59
N	1836			1354

Standard errors in parentheses after the coefficients. * p < 0.05, ** p < 0.01, *** p < 0.001.

presents the results for personal religious change. Importantly, the models had conditioned the fixed effects of parental/premigration religions so that choosing a candidate religion functionally equates to converting to that religion.

3.1. Intergenerational Conversion

There is a positive association between religious traits and intergenerational conversion. We found a significantly positive effect for missionary intensity (0.32, p < 0.05). This number translates to a 38% (e^0.32 – 100%) higher likelihood of conversion when a religion increases its missionary intensity by one standard deviation after fixing paternal religion and personal characteristics. This effect implies that the average rate of missionary intensity determines the utility of choosing a different religion and hence the probability thereof. Additionally, organized activity is also positively associated with changing religion (p = 0.06), which means that for a standard deviation increase in the frequency of organized activities, a respondent would be 63% (e^0.49 – 100%) more likely to convert to that religion. Although the association is not significant at the arbitrary p-value of 0.05, we consider this tempting association evidence for a religious trait effect amid the small number of conversion cases. For the intergenerational religious change from one’s mother’s religion, religious trait effects remain similar to that in the model for one’s father’s religion. There is a significant positive association between missionary intensity and conversion (β = 0.36, p < 0.01) and a positive association between organized activity frequency and conversion (β = 0.49, p = 0.06), after personal characteristics and the mother’s religion were controlled for.

We also found significant associations between several individual-level characteristics and conversion. Women were twice (e^0.69, p < 0.001) as likely as men to choose Buddhism relative to no religion, after we fixed the father’s religion. A similar gender preference for Buddhism exists for maternal religious change. For the age effect, for every 1-year increase in age, the probability of choosing Buddhism drops by about 5% (e^−0.05 – 100%, p < 0.05) after we fixed the father’s religion, and it drops by 6% (e^−0.06 – 100%, p < 0.05) in the mother’s religion model. There is also a migration time effect: for a change from the father’s religion, every 1-year increase in staying brings about a 14% (e^0.13 – 100%, p < 0.01) higher likelihood of choosing Christianity and a 12% (e^0.11 – 100%, p < 0.01) higher likelihood of choosing folk religions, relative to no religion. Migration’s time effect also holds for the mother’s religion model. In addition, every standard deviation increase in financial strain is associated with a 28% (e^0.25 – 100%) higher likelihood of choosing Christianity for the paternal religion model and a 26% higher likelihood for the maternal religion model.

3.2. Personal Conversion

The models in Table 6 analyzed personal conversion by tracking individuals’ current choice of religion to their premigration choices. The coefficients and significance of most variables remained approximately stable compared with the intergenerational models. Model 1, with only religion-specific variables, shows that missionary intensity and religion’s salience were both significantly associated with a higher likelihood of a conversion. After a standard deviation increase in a religion’s missionary intensity, a person’s probability of conversion to that religion raises by 82% ($e^{0.6}-100\%,p<0.01$); after a standard deviation increase in a religion’s collective salience, the chance of converting to that religion raises by 16% ($e^{0.15}-100\%,p<0.001$). When individual-level characteristics were controlled for in the mixed multinomial logit model, shown as Model 2, only missionary intensity was still significantly associated with conversion. Longer stays in the US are significantly associated with choosing Christianity (β = 0.16, p < 0.01) and folk religion (β = 0.19, p < 0.05) over no religion, after premigration religious identification is controlled for. Higher financial strain is also associated with choosing Christianity compared with no religion (β = 0.26, p < 0.05), after premigration religious identification is controlled for.

Interestingly, the entire array of positive and significant fixed effects for premigration religions, with reference to no religion, suggests that the sampled immigrants were more likely to choose a religion than to choose no religion. Furthermore, a comparison of the McFadden R-squares across Table 4, Table 5 and Table 6 indicates that children’s current religious identification is best explained by their premigration religion and less defined by their respective parents’ religions. The divergence comes as not much of a surprise given that the length of stay in the US averages only 3.5 years, rendering the fixed effect of premigration religious identification the strongest predictor of current religious identification.

4. Discussion

This study investigated conversion with a sample of Chinese students in US public universities. By examining both religious organizational traits and personal characteristics, we advance the theory of conversion to one that incorporates factors at two levels that interactively channel the movement of religious identity to a new destination across different candidate choices. Many previous studies on the emerging religious beliefs and practices among Chinese immigrants to North America and Europe have detailed the structural advantage of Christian denominations over other religions (Hall 2006; Wang and Yang 2006; Yang 2010a; Yang et al. 2018). While most of these studies have either employed a qualitative ethnographic approach or only indirectly implied the social mechanism leading to conversion among Chinese immigrants, the current study has empirically demonstrated that individual-level characteristics and organizational factors operate in tandem in conversion. Another merit of the current study is its examination of both personal conversion and intergenerational conversion.

For individual-level characteristics, our mixed multinomial models with no religion as the base reference confirmed that women and younger immigrants were more likely to experience intergenerational conversion toward Buddhism. Longer stays in the US and greater financial strains are associated with favoring Christianity in intergenerational conversions and personal conversions. The acculturation hypothesis proposes that immigrants may adopt the dominant religious system in the host society to construct an assimilated identity for integration and acceptance (Carnes and Yang 2004; Yang 2010a). In this case, Chinese students and scholars who have stayed longer may convert to Christianity during the longer acculturation process. In addition, the opium thesis may explain why the greater financial strain is associated with a higher likelihood of choosing Christianity. A religious organization is filled with human resources, connections, opportunities in other institutions, and material goods for members to partake in and utilize (Maselko et al. 2011; Schnabel 2020). These compensatory resources often enable the participants of organized religious activities to acquire a better socioeconomic status. Chinese students and scholars with weaker family finances to rely on may turn to Christian churches to acquire basic needs, social connections, information, and other resources. The intersectional disadvantage of racial minorities and immigration has imposed on Chinese students and scholars a desire for social status and prestige, which they may compensate for by partaking in the dominant organized religions in the US and exploiting the economic and social opportunities that the religious institution offers.

The current study measured religion-specific organizational traits by aggregating the individual-level responses and using the religion-specific composition of these traits. Such composition effects distinguish an aggregated collective social fact arising from, but irreducible to, individuals’ religious behaviors and practices (Cummins et al. 2007; Duncan et al. 1998), a distinction that has resonated with the notion of faits sociaux since the time of Durkheim and Mauss. We found that although effects slightly varied between intergenerational conversion and personal conversion, a religion with stronger organizational strength consistently attracts incoming members. This study found that the missionary intensity of a religious tradition is significantly associated with a higher likelihood of choosing that religion across all models. In line with the cost–benefit analysis of organizational behavior (Stark and Iannaccone 1994), when a religious denomination mobilizes and sends out proselytizing missions, it may increase the exposure opportunity of its perceived valuable goods to the individuals, with a propensity toward conversion. The aggregated religious salience of religion is also associated with personal religious change, after personal characteristics are controlled for. Given the smaller number of converts compared with other population-level studies, we may also consider a religion’s organized activity intensity at the p-value of 0.06 as evidence for the organizational thesis.

In light of the findings on the organizational correlates of conversion, this study hearkens back to the call for an organizational understanding of religious changes as a function of different religions’ respective capacities to provide rich services for the target population (Glock and Bellah 1976; Stark and Finke 2000; Stark and Iannaccone 1994; Straus 1979; Yang and Tamney 2006). The religions with a stronger recruitment strategy and more missionary supplies prove to possess a higher chance of attracting new believers. Conversion to a specific religion cannot simply come as an individually motivated push toward a new faith system. Even when an individually motivated condition has been satisfied, a multitude of candidate religions exist, and they may appear differentially accessible and appealing to the individual by virtue of their organizational strength. Rational actor theory (Stark and Iannaccone 1994; Stoll and Petersen 2008; Yang 2006) and institutionalism theory (Hall 2006; Straus 1979) concur on the importance of organizational strength in recruiting new members. Conversion, therefore, transpires in a competitive context where religious organizations are endowed with differential capabilities and salience, whose functional importance, in addition to individual-level factors, is indispensable and statistically significant.

The organizational traits of religions also help explain why migrants convert when moving to a different context without necessarily experiencing fundamental changes in individual characteristics. A religion’s organizational force operates within a political and cultural context and can be constrained or promoted in specific contexts. When migrants return to their original country, the reshuffling of relative organizational strength among different religions will also reconfigure the likelihood function of choosing religions, as shown in our mixed multinomial model. In the case of Chinese overseas students who return to mainland China, declining organizational factors of Christianity, in terms of various organized activities, will reduce the attractiveness of Christianity relative to other less-restrained religions or no religion. Therefore, the phenomena of “temporary” Christians or Muslims among Chinese people who are overseas4 should be attributed to the respective religion’s shifting organizational strength, beyond individual conviction or psychosocial motives.

5. Limitations

We acknowledge several limitations in this study. First, this study opens the discussion on conversion among Chinese overseas students and scholars, but readers should restrict the generalizability of the findings to those enrolled in US public universities instead of all overseas Chinese people. Second, the survey behind this study retrospectively asked for premigration information, which may be influenced by recall bias. Future studies should employ a longitudinal panel design when feasible. Finally, the limited number of categories of religion (n = 4) has restricted the analytical power of the conditional logit model. The limited categories of religion imply that any religion-specific trait would have only n − 1 degree of freedom, which has particularly limited our ability to include a variety of religion-specific measures before severe multicollinearity corrupts the model. It has also limited the ways we could have measured the latent utility of the unobservable traits of each religion. Comprehensive research on conversion among immigrants may consider pooling more immigrant populations to allow the inclusion of more religious traditions in the analysis.