Valeria Rainero, Jörg Stolz, Ruud Luijkx
Sociological Science February 10, 2026
10.15195/v13.a7
Abstract
Secularization is one of the most debated areas of research in current sociology of religion. Despite hundreds of empirical studies, researchers do not even agree on the very existence of secularization in different parts of the world. This article investigates whether some of the variability in findings may be attributed not to the social reality investigated but to bias in the form of researchers’ own religiosity. Specifically, we test whether researchers’ religiosity is correlated with two outcomes: their personal belief in the secularization thesis and the likelihood of supporting secularization in their published articles. To address this question, we constructed an international database of scholars working on secularization and conducted a survey measuring their religiosity and beliefs about religious decline. We then coded their publications according to whether they supported the secularization thesis and linked the two data sets. We find significant evidence of a “(non-)religious bias.” Either in their private attitudes or public writings, religious researchers find less evidence for the secularization thesis, whereas secular scholars find more. This result cannot be explained by differences in research methods, study quality, or the religious and geographic contexts under investigation.
Secularization is one of the most debated areas of research in current sociology of religion. Despite hundreds of empirical studies, researchers do not even agree on the very existence of secularization in different parts of the world. This article investigates whether some of the variability in findings may be attributed not to the social reality investigated but to bias in the form of researchers’ own religiosity. Specifically, we test whether researchers’ religiosity is correlated with two outcomes: their personal belief in the secularization thesis and the likelihood of supporting secularization in their published articles. To address this question, we constructed an international database of scholars working on secularization and conducted a survey measuring their religiosity and beliefs about religious decline. We then coded their publications according to whether they supported the secularization thesis and linked the two data sets. We find significant evidence of a “(non-)religious bias.” Either in their private attitudes or public writings, religious researchers find less evidence for the secularization thesis, whereas secular scholars find more. This result cannot be explained by differences in research methods, study quality, or the religious and geographic contexts under investigation.
 | This work is licensed under a Creative Commons Attribution 4.0 International License. |
Supplemental Materials
Reproducibility Package: A replication package with instructions, data, and STATA code is publicly available on the Open Science Framework (OSF): https://osf.io/vcxnk/.
- Citation: Rainero, Valeria, Jörg Stolz, and Ruud Luijkx. 2026. “The Faith Factor. How Scholars’ Religiosity Biases Research Find- ings on Secularization” Sociological Science 13: 154-177.
- Received: October 30, 2025
- Accepted: December 16, 2025
- Editors: Arnout van de Rijt, Andreas Wimmer
- DOI: 10.15195/v13.a7
This is an excellent paper with a strong study design to answer the research question. Congratulations to the authors (RSL afterwards).
We have two remarks: one concerning the empirical results and one concerning the interpretation of the results.
1) Due to the well-documented replication package, we were able to reanalyze the data. The authors’ results were reproduced perfectly. They also proved robust when using a linear probability model instead of logistic regression and implementing no multiple imputation.
However, we noticed that the religiosity effect is nonlinear. A bivariate LOWESS plot reveals a “kink” at a religiosity scale value of 4 (the scale runs from 0 to 8). Rerunning model 4 from Table 4 (outcome: “research conclusion”) with a categorical religiosity variable (eight categories) revealed that religious researchers had an approximately 18 pp lower probability of reporting the finding of secularization in an article. This is similar to what RSL found with their continuous religiosity scale (see their Figure 1). However, the pattern of effects of the religiosity categories differs; it resembles a step function. Non-religious researchers (values 0–4) do not differ much in probability; religious researchers do not differ much either, but they have an 18 pp lower probability.
Thus, the main result of RSL is confirmed, but it appears that the religiosity effect is not gradual, but rather step-like.
2) This raises the question of which mechanisms produce the result. RSL are right to point out that we can only speculate about these mechanisms. However, their study addresses a fundamental issue in the social sciences. Consequently, many readers will speculate about the reasons for the results.
They mention the constructivist position that all research is value-laden. Constructivists may interpret RSL’s Figure 1, which shows a gradual decline in the probability of reporting the finding of secularization, as proof of their position that any researcher is affected by religiosity bias and that objective research is impossible.
RSL also mention the post-positivist position that researchers should strive for value-neutral research, though some may fail in this respect. We speculate that the step-like pattern we found in the data could be interpreted post-positivistically: Most researchers conduct objective research, but some “zealots” distort their research to align it with their values. As RSL mention, we do not know on which side of the religiosity spectrum these zealots are.
According to the post-positivist view, it is crucial to correct the distorted results produced by these zealots. To accomplish this, research must be transparent and open, as RSL mention. Researchers should also use replication packages to reanalyze the data. Instead of publishing one “we-are-the-first” study after the other, sociologists should take each other’s work seriously! In our opinion, this is the only way to produce cumulative knowledge despite religiosity (or ideological) biases among (a few) researchers. RSL should be applauded for making this apparent in their study.