Structural Equivalence of an Attitude Toward Religion Scale in Mexico , Nicaragua and China *

r e S u M e n Las Escalas de Actitudes hacia el Catolicismo, Judaísmo, Hinduismo e Islam (Francis & Enger, 2002; Francis & Katz, 2007; Francis, Santosh, Robbins, & Vij, 2008; Sahin & Francis, 2002) fueron adaptadas en este estudio a una sola escala que evalúa actitudes globales hacia la religión. La escala resultante de Actitudes Hacia la Religión (AHR) fue adaptada a español y chino y fue administrada en México (n=265), Nicaragua (n=296) y China (n=460) a una muestra total de 1021 individuos (59% mujeres, 41% hombres; Medad=22.4 años, DE=7.01 años). La equivalencia estructural de la escala (¿El instrumento mide el mismo constructo en cada país?) fue evaluada utilizando Análisis Factoriales Exploratorios y comparaciones por pares. Los resultados muestran evidencia sólida de equivalencia estructural, ya que se obtuvo una solución unidimensional (a la que se llamó Actitud Hacia la Religión) en los tres países y el valor de Phi de Tucker fue muy cercano a 1. Estos hallazgos apoyan la solución unidimensional de actitudes hacia la religión obtenida por investigaciones previas y amplían el alcance de estos estudios en diferentes contextos culturales. Otras implicaciones también se discuten a profundidad. Palabras clave religiosidad; equivalencia estructural; psicología trans-cultural; psicometría doi: 10.11144/Javeriana.upsy15-2.seat Para citar este artículo: Aguilera Mijares, S., Domínguez Espinosa, A., & Velasco Matus, P.W. (2016). Structural Equivalence of an Attitude Toward Religion Scale in Mexico, Nicaragua and China. Universitas Psychologica, 15(2) 315-320, http://dx.doi.org/10.11144/ Javeriana.upsy15-2.seat * Artículo de investigación. La presente investigación fue posible mediante el financiamiento otorgado a la Dra. Alejandra del Carmen Domínguez Espinosa por el Consejo Nacional de Ciencia y Tecnología CB-2011-1 por el proyecto titulado “El manejo de la impresión y su valor adaptativo desde una perspectiva transcultural” con número de registro CONACyT 168978. ** Correo electrónico: s.am8@hotmail.com *** Directora del Departamento de Psicología Correo electrónico: alejandra.dominguez@ibero.mx **** Departamento de Psicología. Correo electrónico: velasco.matus@gmail.com Santiago aguilera MijareS, alejandra del CarMen doMínguez eSpinoSa, pedro Wolfgang VelaSCo MatuS 316 Un i v e r s i ta s Ps yc h o l o g i c a V. 15 No. 2 a B r i l-j U n io 2016 Religiosity is a complex multidimensional construct. Attempts to assess it have resulted in conflicting models and a wide range of instruments measuring different dimensions of religiosity (Hill & Hood, 1999). In regards with this problem, Francis and Katz (2007) proposed that the attitudinal dimension of religiosity is preferable to others (e.g. affiliation, practice, or beliefs) for constructing an integrated series of studies. The attitudinal dimension is capable of making comparisons both within and between communities, and of appropriate comparisons across wide age ranges. Besides, it is more stable and deep-seated within an individual, less likely to be contaminated by personal or contextual factors, and adequately assessed by Likert scales (Francis et al., 2008). As well as being multidimensional, religiosity is multifaceted in the sense of including many different doctrines (e.g. Christianity, Hinduism, etc.). Francis (1978) argued that the attitudinal dimension of religion can be best assessed through the specific tradition by which it is expressed. He thus created a scale measuring attitude toward Christianity and, later, one for attitudes toward Judaism (Francis & Katz, 2007), towards Islam (Francis et al., 2008), and towards Hinduism (Sahin & Francis, 2002). We merged these scales into a single one in order to test the liability of a general scale of attitude toward religion (i.e. not doctrine-specific) which would render empirical cross-cultural studies of religiosity possible. Cross-cultural studies are non-experimental in nature, since participants cannot be randomly assigned to cultures, nor can the compared groups be matched on background variables (Van de Vijver & Leung, 2011). As a consequence, comparing measurements across cultural groups is prone to bias. Van de Vijver and Leung (2011) therefore argued that a main concern of cross-cultural research should be maximizing the validity of its inferences, and that establishing cross-cultural equivalence and suppressing bias are crucial to such task. The main objective of the present study is to establish whether the Attitude toward Religion Scale has equivalence of measure at a structural level. In other words, the goal is to find whether the construct measured by this scale is the same across the cultural groups studied, or whether it only overlaps partially. Evidence for such equivalence was assessed across a group from Mexico, one from Nicaragua, and another from China. In Mexico, 82.7% of the population is Catholic, 9.7% is protestant, and 4.7% is atheist (Instituto Nacional de Estadística y Geografía, 2010). As for Nicaragua, 73% of the population is Catholic, while 25% is protestant, and 8.4% is atheist (Instituto Nacional de Información de Desarrollo, 2005). China has no nationally representative surveys regarding the religious affiliation of its population. However, a survey (The Pew Global Attitudes Project and the Committee of 100, 2007) carried out in urban populations reports 14% of Chinese adults are religiously affiliated (12% Buddhist, 1% Catholic, 1% Christian, < 1% Muslim, < 1% Taoist). Such demographical data suggests that Mexico and Nicaragua may have similar religious backgrounds, whereas China’s is markedly dissimilar. Comparing samples from these three groups is thus relevant for assessing the universality of the attitude toward religion construct. Given the satisfactory psychometric properties of L. J. Francis’ doctrine-specific measures of religiosity (Francis & Enger, 2002), our hypothesis is that the Attitude toward Religion Scale will achieve to provide evidence of structural equivalence across the three cultural groups studied.

Religiosity is a complex multidimensional construct.Attempts to assess it have resulted in conflicting models and a wide range of instruments measuring different dimensions of religiosity (Hill & Hood, 1999).In regards with this problem, Francis and Katz (2007) proposed that the attitudinal dimension of religiosity is preferable to others (e.g.affiliation, practice, or beliefs) for constructing an integrated series of studies.The attitudinal dimension is capable of making comparisons both within and between communities, and of appropriate comparisons across wide age ranges.Besides, it is more stable and deep-seated within an individual, less likely to be contaminated by personal or contextual factors, and adequately assessed by Likert scales (Francis et al., 2008).
As well as being multidimensional, religiosity is multifaceted in the sense of including many different doctrines (e.g.Christianity, Hinduism, etc.).Francis (1978) argued that the attitudinal dimension of religion can be best assessed through the specific tradition by which it is expressed.He thus created a scale measuring attitude toward Christianity and, later, one for attitudes toward Judaism (Francis & Katz, 2007), towards Islam (Francis et al., 2008), and towards Hinduism (Sahin & Francis, 2002).We merged these scales into a single one in order to test the liability of a general scale of attitude toward religion (i.e.not doctrine-specific) which would render empirical cross-cultural studies of religiosity possible.
Cross-cultural studies are non-experimental in nature, since participants cannot be randomly assigned to cultures, nor can the compared groups be matched on background variables (Van de Vijver & Leung, 2011).As a consequence, comparing measurements across cultural groups is prone to bias.Van de Vijver and Leung (2011) therefore argued that a main concern of cross-cultural research should be maximizing the validity of its inferences, and that establishing cross-cultural equivalence and suppressing bias are crucial to such task.
The main objective of the present study is to establish whether the Attitude toward Religion Scale has equivalence of measure at a structural level.In other words, the goal is to find whether the construct measured by this scale is the same across the cultural groups studied, or whether it only overlaps partially.
Evidence for such equivalence was assessed across a group from Mexico, one from Nicaragua, and another from China.In Mexico, 82.7% of the population is Catholic, 9.7% is protestant, and 4.7% is atheist (Instituto Nacional de Estadística y Geografía, 2010).As for Nicaragua, 73% of the population is Catholic, while 25% is protestant, and 8.4% is atheist (Instituto Nacional de Información de Desarrollo, 2005).China has no nationally representative surveys regarding the religious affiliation of its population.However, a survey (The Pew Global Attitudes Project and the Committee of 100, 2007) carried out in urban populations reports 14% of Chinese adults are religiously affiliated (12% Buddhist, 1% Catholic, 1% Christian, < 1% Muslim, < 1% Taoist).Such demographical data suggests that Mexico and Nicaragua may have similar religious backgrounds, whereas China's is markedly dissimilar.Comparing samples from these three groups is thus relevant for assessing the universality of the attitude toward religion construct.
Given the satisfactory psychometric properties of L. J. Francis' doctrine-specific measures of religiosity (Francis & Enger, 2002), our hypothesis is that the Attitude toward Religion Scale will achieve to provide evidence of structural equivalence across the three cultural groups studied.

Instruments
Attitude toward Religion Scale.Adapted from the Attitude toward Christianity, Judaism, Hinduism, and Islam scales (Francis & Enger, 2002;Francis & Katz, 2007;Francis et al., 2008;Sahin & Francis, 2002).Based on common dimensions from previous scales (e.g.God, bible, prayers, church, synagogues, religion, church attendance, among others), the Attitude toward Religion Scale consists of 17 items that evaluate an affective response toward God (e.g.God helps me carry on a better life), toward religious practices or symbols (e.g.I think the religious scripts are antiquated), and toward religious beliefs in general (e.g.My religious beliefs truly shape the global scope of my life).The scale was structured in a five point Likert scale format (1 = strongly disagree to 5 = strongly agree), and was adapted to both Spanish and Chinese using the back-translation method proposed by Brislin (1970) and Willgerodt, Kataoka-Yahiro, Kim, and Ceria (2005).Francis and Enger (2002), Francis and Katz (2007), Francis et al. (2008), and Sahin and Francis (2002) have previously reported good psychometric properties for each of their unidimensional scales, with Cronbach's Alpha values above 0.85 and within a range of 30-60% of total explained variance.

Procedure
With an auto-applicable format, the instrument was handed out throughout schools, work offices and homes, either in individual or group sessions.It was administered within a battery containing other psychological scales, intended for different studies.
The respondents agreed to participate voluntarily, anonymously, and received no economical compensation for doing so.The average time for completion was approximately 10 minutes.

Statistical Analyses
All analyses were conducted using the SPSS statistical program (2010).The descriptive statistics and reliability coefficients were calculated in order to gain insights regarding the components of the instrument.As for assessing structural equivalence, Exploratory Factor Analyses (EFA) and the Tucker's congruence coefficient were conducted.
Structures found in the factorial matrices were target rotated and Tucker's congruence coefficients per scale and per factor were carried out to evaluate their similarity by pairwise comparisons.

Results
The descriptive statistics and Cronbach's Alpha (a) are shown in Table 1.The mean of the full samples was 55.49 (SD= 15.77).A one-way ANOVA was conducted to test the differences between the group's means and its results show a statistically significant difference (F [2, 949] = 304.55;p < 0), being Nicaragua the highest (M = 6.6,SD = 13.31) and China de lowest (M = 46.28,SD = 10.6).
The Alpha coefficient for the total sample was 0.94.According to Cohen (1988), J. Gliem and R. Gliem (2003), and George and Mallery (2003), reliability coefficients of 0.60 or higher are satisfactory for comparisons within groups.All three group's alpha coefficients were therefore satisfactory.
Table 2 shows the factor loadings for the EFA carried out.The EFA was fixed to look for one fac- tor, because the attitudinal dimension of religiosity is unidimensional.The group's factor analyses also need to have the same number of factors in order to be target rotated and subjected to the Tucker's congruence test.The results of the EFA support the decision to fix for one factor, since all but three factor loadings (items 9, 12, and 15 for China's group) were under 0.40 and therefore non-significant.
In regards with Tucker's congruence coefficient, as seen on Table 3, the values ranged from 0.98 to 0.99.Values that are higher than 0.95 indicate factorial similarity, and values lower than 0.90 correspond to non-negligible differences in factorial structure (Lorenzo-Seva & ten Berge, 2005;Van de Vijver & Poortinga, 1994).This means the analyses' results provide strong evidence of structural equivalence for the ATR scale.Note: All items were adapted from the Attitude toward Christianity (Francis & Enger, 2002), Attitude toward Judaism (Francis & Katz, 2007), Attitude toward Hinduism (Francis et al., 2008), and the Attitude toward Islam (Sahin & Francis, 2002) into the ATR Scale in order to measure overall attitudes toward religion.
Source: Own work

Conclusion
In conclusion, the Attitude toward Religion Scale that we built from previous scales can be regarded as a structurally equivalent measure of religiosity across cultures as different as Mexico, Nicaragua, and China.This implies that the construct of religiosity that the ATR scale targets appears to be universal, and cross-cultural.This is important because in cross-cultural studies the construct measured must be equivalent for the comparison to be meaningful (Van de Vijver & Poortinga, 1997; Van de Vijver & Tanzer, 2004), and specifically about religious assessments, measures have remained culturally specific (Francis & Katz, 2007); therefore, our findings broaden the scope for religious studies.Francis (1978Francis ( , 2009) ) argued that the attitudinal dimension of religiosity is best measured using scales specific to the doctrine in which it is expressed, and she therefore created different scales toward different religions.Our goal was to develop a scale that evaluated an overall attitude towards religion, and although it was developed considering Mexico and Nicaragua's Catholic background, our construct remained stable even in a totally different religious-cultural context such as China.Our findings propose that this construct of Attitude Toward Religion may be adequately assessed throughout diverse cultures without the need for the measuring instrument to be adapted to the specific religious doctrine of the studied group.
Our psychometric evidence shows that the ATR Scale maintains certain stability within different cultural contexts; these psychometric properties enhance the scale's usefulness and its attractiveness to students and scholars worldwide, as suggested by Karam, Sekaja, and Geldenhuys (2016).However, although Tucker's congruence coefficient provided evidence for the structural equivalence of the instrument, it is important to note that before valid cross-cultural comparisons can be made, this instrument needs to be tested for the highest level of equivalence: scalar equivalence.Only then can the scores obtained in different cultures be directly compared with one another.
Structural EquivalEncE of an attitudE toward rEligion ScalE in MExico, nicaragua and china Un i v e r s i ta s P s yc h o l o g i c a V. 15 No. 2 a B r i l -j U n i o 2016 317

table 1
Descriptive statistics and Cronbach's Alpa for the ATR Scale Note: ATR= Attitude toward Religion; a = Cronbach Alpha.Source: own work

table 2
Exploratory Factor Analysis for the Attitude Toward Religion (ATR) Scale, separated by group