It is not surprising that South Dakota and Kentucky have higher fertility rates than Rhode Island and California. People in the former states tend to be rural, religious, and conservative. People in the latter states tend to be highly urbanized, secular, and liberal. It is also intuitive that these characteristics are correlated: one can easily imagine a god-fearing, Trump-supporting farmer and a cosmopolitan socialist who has not been to church in years. The question I am curious about is which one of these variables has the strongest effect on birth rates.
Data for political patterns and urbanization levels was readily available, but was lacking for religion. The Census Bureau’s Household Pulse Survey includes a question on attendance of religious services at least four times a year, which I used to estimate the religiosity levels of each state (specifically phases 4 to 4.2). Although imprecise, it yielded the highest explanatory results on birth rates.
A 10% increase in religious attendance is associated with 3.4 more births per 1000 women. A 10% increase in Republican voter share (using the Cook Partisan Voting Index) corresponds to 2 more births. After controlling for religious and political tendencies, the level of urbanization was not meaningfully associated with the birth rate. Rural states had some of the highest and lowest fertility rates.
The r squared value being above 0.6 is a strong result for modeling complex human behavior using only three variables. In other words, over 60% of the differences in states’ fertility rate can be explained by the differences in states’ religiosity, partisanship, and urbanization alone (although these are very important and encompassing variables). The log-transformed model has an even higher positive correlation. Despite the correlation between religiosity, conservatism, and a larger rural populace, the variance inflation factor was less than 5 for all the variables, which is a common threshold for ensuring the interconnectedness of independent variables is not too high to affect a model.
Using the standard (not logarithmic) model, I calculated the expected fertility rate in each state. I was happy with how similar the maps looked, with the regional trends being very apparent.
These results are interesting to me because they challenge certain prevalent ideas about American society. People discuss political polarization as a monolithic force, with the growing divide between a red and blue America. History tells us this is not a new phenomenon (the Civil War, 1890s populist movement, 1960s/70s counterculture, etc.), even if it has exacerbated in recent decades. However, religious attendance bearing more explanatory power in a behavior as important as having children than political affiliation suggests a flaw in contemporary discourse. Ideas about religion and philosophy, which often are traced back thousands of years, reflect something important about the human condition. Your personal beliefs, as well as those in your social sphere, on these topics probably reflect your worldview more than your opinion on the president’s latest controversial comments.
The results also contest other widely held superficialities in 2020s America. The regions with the highest and lowest fertility rates, the Great Plains and New England respectively, both contain the highest proportion of white residents in the country. Our society obsesses over race as a monolithic concept to explain people’s behavior, when race is often better understood as a conduit for more meaningful social or cultural affiliations. Suggesting similarities solely on racial grounds is condescending, white farmers in Vermont and South Dakota stand opposite on almost any social metric. People are dynamic forces, not machine-like products of their environments.
From that final point, it would be misinterpreting the data to think there would be exactly 3.4 more kids per 1000 women if 10% more Americans started going to church. The main value from this information is its suggestion that religion plays a more important part than political affiliation or rural residency in predicting the decision to have kids. This raises interesting questions about human psychology and cultural dynamics.
Sources:
Cook Partisan Voting Index (https://www.cookpolitical.com/cook-pvi/2025-partisan-voting-index/state-map-and-list)
US Census dataset on State-Level Urban and Rural information for the 2020 Census and 2010 Census (https://www.census.gov/programs-surveys/geography/guidance/geo-areas/urban-rural.html)
CDC National Center for Health Statistics (https://www.cdc.gov/nchs/pressroom/sosmap/fertility_rate/fertility_rates.htm)
Household Pulse Survey dataset on Lack of Social Connection (https://www.cdc.gov/nchs/covid19/pulse/lack-socialconnection.htm)
© Peter Derrah 2025. All rights reserved.