Details about the Muslim and non-Muslim models

The model fits as well in each sample (e.g., NFI = .983 for the Muslim sample and .993 for the non-Muslim sample); the seven questions each significantly relate to the common factor religious-cultural negativity (RCN) (P < .001, two-tailed) and load onto the factor at .50 or stronger. There were only minor differences in strength. We also tested the model at the country level, and found that the RCN factor fit the data as strongly or more strongly—for example, among Nigerian Muslims (RMSEA=.029), Nigerian non-Muslims (RMSEA =.035), in Turkey (RMSEA =.031), among French Muslims (RMSEA=.023), and French non-Muslims (RMSEA=.000). While it would be possible to use the regression coefficients or factor loadings generated in the factor analyses to adjust the index to match how the seven questions relate to the common factor on a country-by-country basis, we do not do this because in most cases the differences are small. Moreover, to do so would distort the true relationships in the data by inflating the value of variables in some countries, depending on which endogenous variable is set to one, which is an arbitrary choice in factor analysis.