This study compares the estimates of religious composition of counties in the United States from three independent datasets: the 2010 Religious Congregational Membership Study (RCMS); the 2010 Infogroup Congregational membership data (INFO); and the Gallup Daily Poll—a large national survey with more than 1.3 million respondents. My analyses suggest that the estimates for most major religious groups from the three datasets are highly correlated to each other. In addition, the measures of local religious compositions from the three datasets successfully predict the religious composition of friendship networks in a large, nationally representative survey. These findings suggest that RCMS, the most widely used data source for measuring local religious composition in the United States, has a convergent and predictive validity. My analyses, however, also highlight important challenges in measuring geographic distributions of non‐Christian populations, as well as total religious populations in all religious traditions.