We model the effect of Protestant vs. Catholic denomination in an economic theory of suicide, accounting for differences in religious-community integration, views about man's impact on God's grace, and the possibility of confessing sins. We test the theory using a unique micro-regional dataset of 452 counties in 19th-century Prussia, when religiousness was still pervasive. Our instrumental-variable model exploits the concentric dispersion of Protestantism around Wittenberg to circumvent selectivity bias. Protestantism had a substantial positive effect on suicide in 1816-21 and 1869-71. We address issues of bias from mental illness, misreporting, weather conditions, within-county heterogeneity, religious concentration, and gender composition.