Part 4 – Religion in America: Evidence from Cellphone Data
In this post, I compare the frequency of religious worship attendance using my cellphone data with self-reported attendance separately by religion.
Protestants: Using cellphone data, I classify 35.9% of Americans as Protestants who attend at least once. One out of 15 Protestants attend “weekly” (at least 36 weeks during my 47-week sample).
Catholics: Using cellphone data, I classify 8.7% of Americans as Catholics who attend at least once. One out of 51 Catholics attend “weekly” (at least 36 weeks during my 47-week sample).
Latter-day Saints (Mormons): Using cellphone data, I classify 1.4% of Americans as Latter-day Saints who attend at least once. One out of 7 Latter-day Saints attend “weekly” (at least 36 weeks during my 47-week sample).
Jehovah’s Witnesses: Using cellphone data, I classify 0.44% of Americans as Jehovah’s Witnesses who attend at least once. One out of 15 Jehovah’s Witnesses attend “weekly” (at least 36 weeks during my 47-week sample).
Orthodox Christians: Using cellphone data, I classify 0.41% of Americans as Orthodox Christian who attend at least once. One out of 46 Orthodox Christians attend “weekly” (at least 36 weeks during my 47-week sample).
Jews: Using cellphone data, I classify 0.74% of Americans as Jews who attend at least once. One out of 247 Jews attend “weekly” (at least 36 weeks during my 47-week sample). [CAREFUL: Many observant Jews may not have cellphones during service]
Muslims: Using cellphone data, I classify 0.29% of Americans as Muslims who attend at least once. One out of 58 Muslims attend “weekly” (at least 36 weeks during my 47-week sample).
Buddhists: Using cellphone data, I classify 0.31% of Americans as Buddhist who attend at least once. One out of 775 Buddhists attend “weekly” (at least 36 weeks during my 47-week sample).
Some overall takeaways: 1. I find evidence in all religions that people over report the frequency of religious worship. 2. The number of “weekly” attenders in all religions is small compared to the number of people who attend less regularly.
3. There are differences in the fraction of “weekly” attenders by religion. For example, 14x more Americans identify as Catholics than Latter-day Saints (Mormons), but slightly more weekly-attending Latter-day Saints than Catholics (.20% of Americans compared to 0.17%).
3 cont. But because there are so many Catholics compared to Latter-day Saints who attend once a month or a few times a year, there are far more Catholics in church on any given Sunday (~4M compared to ~1.4M).
4. The vast majority of “weekly” attenders in the US are Protestants (85%).
• • •
Missing some Tweet in this thread? You can try to
force a refresh
Part 7 - Religion in America: Evidence from Cellphone Data
In this post, I present histograms for the amount of time spent at places of worship on the primary day of worship for each religion.
Part 6 - Religion in America: Evidence from Cellphone Data
In this post, I present histograms by religion for the time of day when people arrive at their place of worship on the primary day of worship.
Part 5 – Religion in America: Evidence from Cellphone Data
In this post, I show the geographic variation in religious attendance based on cellphone data. Here is a state-level graph that shades by fraction of people in each state who attend at least once:
Here is the same graph at the county level:
Getting county-level measures of religious attendance from surveys is *very* hard given sample size issues. Even in my sample of >2M, estimates can become noisy in small counties.
Part 3 – Religion in America: Evidence from Cellphone Data
I compare frequency of religious worship attendance using my cellphone data with self-reported attendance data. Here are the main results:
According to self-reported data, 66% of Americans affiliate with a religion and 46% claim to attend worship services at least once a year. This matches up nicely with my cellphone data that finds that 48% of people attend at least one service each year.
30% claim to go at least once a month, but I find that only 13% attend at least 11 weeks during the 47 weeks of my data sample.
Part 2 – Religion in America: Evidence from Cellphone Data
Based on the 2.1M cellphones in my main sample that I discussed in “Part 1”, here is the estimated number of worshippers in the US each week from April 2019 to Feb 2020:
~30 million Americans attend a worship service each week. Attendance increases by ~50% for Easter and Christmas. There are dips in attendance on holiday weekends.
In the next 8 figures, I show attendance separately for the 8 largest religions (using Pew's classification).
Protestants: ~22M in attendance each week. Large Easter and Christmas spikes.
Part 1 – Religion in America: Evidence from Cellphone Data
Sorry, but this first post has to be boring. I need to tell you about the 2.1M cellphones that make up my main sample.
The cellphone data were purchased from a company called @veraset that aggregates de-identified, geospatial data points for US smartphones. It’s a bit of a blackbox, but if you allow apps like the weather app to track your location, you could be in their dataset.
The most detailed data contain “pings” (lat-longs) for the location of each phone. A regularly pinging phone will produce a ping every 5 minutes or more frequently. Using polygon shape files (geofences) of buildings, you can tell when a phone enters a Walmart, Subway, or church.