Religious attendance and global cognitive function: A fixed-effects cross-lagged panel modeling study of older U.S....

Evidence linking religious attendance to better cognitive function is based on flawed study designs. No population representative longitudinal studies on the topic have taken both unobserved confounding and reverse causation into account. Recently developed fixed-effects cross-lagged panel modeling (FE-CLPM) offers simultaneous traction on these issues. It also allows parsing of long-from short-run effects. Using FE-CLPM and ten-year data from the Health and Retirement Study-a national probability sample of U.S. adults over age 50-this study began to fill the gaps above.

Gender-specific FE-CLPM models were used to examine bidirectional and within-person linkages of religious attendance with global cognitive function. Granger-Sims "causality" tests further examined short-run effects in both directions. Impulse response analysis was used to explore time patterns in these linkages.

At least among women, religious attendance had negative short-term prospective linkages with global cognitive function. Over successive time points, these associations increased in strength among both genders. Feedback effects-of cognitive status on religious attendance-were found in both women's and men's models, but had a gender-specific pattern.

Results contradict a large literature positing cognitive benefits of religiosity. Instead, they lend support to a recent "neural resource depletion" model-especially among women. Overall, findings illustrate the "dark side" of religious engagement, which studies increasingly present as a social determinant with "outcome wide" positive effects on multiple health dimensions.