Ebbs and flows in American religiosity
When most people think about religion in the United States, they likely imagine a Catholic or Protestant church. Some will think of other Christian denominations, or even a neighborhood synagogue or mosque. A few will conjure a foreboding image of the early Puritan days of the American colonies. But on the ground, religiosity in the United States is much more complex than these snapshots can capture.
To study religion, our Modeling Religious Change project has developed five dimensions of religiosity: religious self-identification, public religious practice, private religious practice, supernatural worldviews, and subjective importance of religion. These dimensions aim to capture multiple facets of religious life and build a better understanding of what religion has historically looked like in the United States.
Once we understand the past, we can use this data to imagine how religious groups will change in the future.
In 1963, Jesuit sociologist John L. Thomas wrote that “On the surface, at least, religion in American history appears to be characterized by a rhythmic ebb and flow of vitality. The last great wave, popularly associated with the name of revivalist Billy Graham, stimulated numerous observers to offer their interpretations of what they assumed to be a revival of religion, or of interest in religion, during the past two decades.”
Thomas published his book, Religion and the American People, during the mid-twentieth century, a period associated with an upswing in American religiosity and religious pluralism. Americans flocked to evangelist rallies, the National Prayer Breakfast began under Dwight Eisenhower, and Congress adopted ‘In God We Trust’ as the official motto of the United States.
The 1950s and early 1960s encapsulated in Religion and the American People constituted a rise in religion. In a cyclic model of religion, a fall in religiosity would come next.
But religious history is complicated.
Secularization theory suggests that Americans became much less religious during the second half of the twentieth century, and this trend will continue into the foreseeable future. Yet the data on secularization is conflicted, leading some social scientists to outright reject the idea that American religiosity will keep falling.
So, what did happen to American religiosity in the late twentieth century? And what is happening now?
Spotlight on religious attendance
Religious attendance is a key component of public religious practice, a term that refers to how regularly an individual engages with a religious community. Surveys and polls typically ask about religious attendance, which makes it one of the most important and widely used metrics for studying public religious practice. In the United States, data on attendance from these surveys and polls goes all the way back to the late 1930s. Can we rely on these data?
Several elements complicate the accurate tracking of religious attendance.
- One issue is the language surveys use. A survey might use terms like “frequently” or “seldom,” which are subjective and don’t have an inherent numerical value. A survey might also ask about religious attendance but exclude attendance for specific reasons like weddings or holidays.
- The number and nature of responses that surveyed individuals could select can also vary from survey to survey. Participants may be given as few as four responses to choose from (never, seldom, occasionally, frequently), while other surveys are open-ended and participants are not restricted in their answer.
- It is well-documented that Americans tend to over-estimate their religious attendance. Social desirability can lead individuals to over report their attendance if a higher level of religious attendance is considered culturally admirable. During the Cold War, for instance, it was important for Americans to be religious — distinguishing themselves from the Communist Atheists in the Soviet Union.
- The survey design may exclude individuals, which leads to them being under-represented. For example, political polls most often survey white men of voting age, as this demographic is the most likely to vote.
- Historical data and its technical information can be lost! For example, after the data collector died, the datasets Thomas used were lost. All that remains of this seminal wide-scale quantitative measurement is the data Thomas had already published in his book.
- Additionally, some religious communities are so small that it is difficult (or even impossible) to accurately survey them. For instance, Jews are tiny minority at 2% of the contemporary American population; Muslims are an even smaller minority at 1%. Other religious communities like Hindus, Sikhs, and Wiccans are smaller still. For these groups, the available datasets are simply not wide enough to allow for the sort of analysis that social scientists can do for Protestants or Catholics.
All this being the case, how do social scientists handle the variation in attendance measures to observe trends across surveys with differing characteristics?
One approach is to place all the data on a similar scale, and then examine how the pattern differs by survey-specific factors. A common scale for attendance, for instance, could be the times per year the person attends. If they attend on a weekly basis, that’s about fifty-two times a year.
The tricky part is interpreting subjective, non-numerical language present in some of the survey questions. The meaning of attending church “regularly” changes over time, for instance, and people in a religious society may feel pressured to pick a higher response if they are forced to choose between responses like “often” and “seldom.”
Our team has compiled data on religious attendance across 147 surveys and polls, stretching back to 1939. We standardized the responses on a common scale, assigning phrases like “frequently” as if they were weekly, and “often” as if they were monthly. This can work, but it is still important to recognize that little variations between surveys can result in very different results.
Let’s look at some examples.
Below are two graphs. The first shows what percent of people in the survey said they attended every week, with the second showing the percent who attended three times a month or more often. In both graphs, we see the overall trend of increasing attendance in the 1950s and 1960s followed by decline, as we might expect.
The first graph, however, appears less consistent than the second. The historical 1955 data points (in pink) and many of the datasets collected by Pew Research Center (in light blue, yellow, and purple) are clear outliers. The 1955 estimates have as few as 15% or as many as 50% of people attended church weekly or more!
Estimates of weekly attendance in Pew surveys is about 15–20 percentage points higher than the General Social Surveys and American National Election Surveys. What accounts for these gaps?
Examining the survey from 1955, we learn that individuals were actually asked about their attendance three times.
- First, they were limited to answering with the following vague responses: regularly, occasionally, seldom, or never. In our analysis, frequently is converted to the same value as weekly, and occasionally becomes monthly. The estimate of weekly and nearly weekly attendance from that question stays the same in both graphs.
- The respondents were also asked how many times they attended in the past year, but the responses are in broad categories like 35–44 and 45–54 times. We quantify these responses as 39.5 and 49.5 times per year, respectively. Both responses fall below the definition of weekly (52 times per year) but above nearly weekly (36 times per year). In comparing the graphs, the two data points align perfectly with the nearly weekly definition.
- Finally, participants in the 1955 survey were asked how many times they attended church in the past month. Attending four times a month or more (48 times per year) is below the weekly threshold, but three times a month aligns with nearly weekly. In the second graph, the responses to attendance in the past month are only slightly lower than attendance from questions about the past year.
Taken together, it seems that people in the mid-50s may consider “frequent” church attendance as going “nearly weekly” instead of “weekly.” We must carefully consider how people would interpret the survey question and what kind of response options were available.
In comparing the Pew surveys to other national surveys, we can see that Americans tend to round up their answers about religious attendance. Pew surveys allow participants to choose from seven responses, with once or twice a month and once a week next to each other. It is not possible to select something in between, so people may select once a week even if they don’t attend quite that often. The General Social Surveys, however, include about once a month, nearly weekly, and once a week as possible responses.
In the first graph, Pew’s estimates of weekly attendance are consistently higher, but in the second graph, the gap between Pew and other surveys’ estimates of nearly weekly are much closer.
Knowledge of the ways survey-specific factors can influence the numbers we see can help us interpret them carefully. We might, for instance, have more confidence that the second graph captures trends in frequent church attendance fairly well, but we are less certain about trends in weekly attendance specifically.
Let’s go one step further to compare trends between religious groups. The third graph shows estimates of nearly weekly attendance that are separated by respondents who identify as Catholic, Protestant, or Unaffiliated.
Even with inconsistencies between survey estimates, a relatively clear pattern emerges. Catholics appear to attend services less frequently, with a sharp decline occurring in the 1970s, while Protestants appear to steadily increase their frequent attendance. Dissent over the Catholic Church’s stance on birth control in the late 1960s may explain the decline in Catholic religious attendance when oral contraceptives were becoming more widely available to American women.
Perhaps the decline in attendance is primarily a Catholic phenomenon. It is important to consider, however, that the Unaffiliated population in the U.S. continues to grow rapidly, with Protestants disaffiliating faster than Catholics. Protestants may appear to be more religious, but former Protestants who were less religious are now answering surveys as Unaffiliated, and their attendance is relatively low.
How MRC studies change over time
Modeling Religious Change is an interdisciplinary project that brings together demography, religious studies, and computational models to engineer simulations of religious change over time. While this simulation is the end result, many elements go into its creation.
By carefully compiling data from many different historical and contemporary sources, we have a clearer picture of the past American religious landscape and consider how it will look in the future. Historical polls and surveys are crucial because they allow us to test how accurate our model is with real-world data. If the model confirms patterns that we know are true about the past, that demonstrates that its projections about the future are on track too.
We know that American religiosity generally increased during the 1950s and early 1960s, and that American Catholic religious attendance dropped during the late 1960s and early 1970s.
But how will American religion change in the future?
That question is one with vital real-world implications. Civic planners need to take the evolving American religious landscape into account to best meet the needs of the American population. Religious changes influence what locations new immigrants move to, where new schools and religious buildings need to be built, and where conflicts based on religious tension might arise.
Religious organizations can also be sites for disseminating public benefits. Churches, for example, can serves as locations for medical intervention. To ensure that these benefits reach the individuals who need them, organizers must understand how frequently individuals engage with those religious organizations — or even if another site would be more useful. Accurate data on religious attendance has deep civic importance.
By studying religious attendance in the United States, we gain new insight into the past, present, and future of American religiosity. By studying how religious attendance has changed in the past, we better understand how the American religious landscape will continue to evolve. In turn, this knowledge will help us to better plan for the present, and for the years to come.
