Religion and Ethnicity: Paradoxes and Scientific Challenges

Abstract

This study investigates connections between religion and ethnicity based on the ISSP Religion’18 module’s data , collected in 2018/19 in 28 countries from all over the world. The focus is on individual religiosity and ethnic self-identities in a general context, personally reported by 39,115 respondents. Implementing a purposefully designed algorithm the societies have been split up into majority and minority ethnic groups, and statistical modeling was used to determine the influencing factors for their levels of religiosity. The key results of the analysis showed a significant division between religiosity, belonging to a denomination, and believing in God. These three religious components had different impacts on ethnic identity - belonging standing out as the strongest one. Ethnic minorities (not only migrants) have a higher level of religiosity, believing, and belonging, compared to ethnic majorities. Believing in God does not necessarily mean believing in religious markers like life after death, heaven, hell etc., which could be considered both as a paradox and as a scientific challenge. Furthermore, religious and ethnic identities have hybrid characteristics and depend on cultural and economic environment - GDP registered a high correlation with religiosity.

Keywords:

• religiosity
• belonging
• believing
• majority vs. minority ethnic groups
• random forest statistical modeling

Notes

1 Detailed description and additional information about ISSP could be found on http://www.issp.org

2 The list of the countries could be seen in Table 1 in the results section.

3 Despites UK and USA successfully completed the ISSP ’Religion module, their teams measure ethnicity in ways, that do not allow to implement our unified algorithm to divide their societies into majority and minority ethnic groups. That is why these countries are not included in our analyses.

4 For Christianity we considered all varieties reported by the respondents - Catholicism, Orthodoxy, Protestantism, other Christian religions. By Asia/Easter religions we meant Buddhism, Hindu and all others indicated by the surveyed individuals.

5 We presume that in every ISSP member country there must exist an ethnic majority cohort and at least one minority community, which is different in some way from the majority. Furthermore, it was a challenge to discover similarities and differences between the two major strata in the context of religion and ethnicity interconnection. Working with the two selected ethnic groups, we were able to compare the main social markers of both religiosity and ethnic identity.

6 We used ‘country’s political past’ as a criterion for grouping the surveyed courtiers into Western European Countries, Post-Socialist European Countries and Overseas countries. Thus, we tried to measure the influence of socialist and non-socialist political regimens on religiosity and mainly to discover what is the echo of Marxist ideology that ‘religion is an opium of the people’ in today's post-socialist societies. Of course, we did NOT place under a common denominator all post-socialist countries (in our case they are only 8), neither in terms of their religious heterogeneity nor in terms of specific dominant and non-dominant religions.

7 Within the group of the 8 surveyed post-socialist countries, Catholic Christianity dominated in 5 of them and Orthodox Christianity in Bulgaria, Russia, and Georgia – all of them, as it is in all other surveyed societies, are religious heterogeneous and the dominant religion does not mean that it is the only existing religion.

8 The method has multiple advantages that clearly leverage the insights and the predictive power of the data. Among these are: the non-linear approach to explain variance in dependent variable, the interactivity in building multiple independent models, the lesser risk of overfitting and generally the highest level of accuracy. One of the most effective advantages is the possibility to extract the variable importance. For this particular task no prediction on target variable was used. The purpose of running the model was to measure the actual importance of the individual drivers in the context of their mutual dependence and interaction with the other selected variables. We focus on Interactive tuning of hyperparameters, that was made using 10-fold cross validation greedy algorithm to determine: the number of variables randomly sampled as candidates at each split (mtry) and the minimum number of data points for further splits in a tree. (min_n) (Breiman 2001). Other hyperparameters were set as constant i.e., number of trees = 1000.

9 Here we used more socio-demographic variable as compared to the previous analysis including top-bottom 10-point scale subjective for respondent’s self-assessment of where he/she stands in society where 0 corresponds to society bottom, and 10 - the highest levels of society.

10 Here we deliberately refrained from using standard Linear Regression (OLS method) due to its multiple limitations and dubious results with categorical data.

11 GDP per capita is gross domestic product divided by midyear population (Source: World Bank)

Additional information

Notes on contributors

Lilia Dimova

Lilia Dimova is a social scientist, Director of the Think Tank organization Agency for Social Analyses (ASA), established in Bulgaria. Assoc. Prof Dimova is the ISSP Principal Investigator (1994 since) and the ESS National Coordinator (2005–2019). Her main scientific interests are in the field of social stratification and inequalities (gender, ethnic, generational, religious), poverty, exclusion/inclusion, ethnic and national identities.

Martin Dimov

Martin Dimov is a data scientist, modeller and statistician, Director of the Data Science Team in Gem Seek Company. Dr. Dimov’s professional achievements are in the field of development of predictive models, innovative approaches and techniques for better utilization of internal data, implementing top-notch ML/AI algorithms in mega and social data analyses.