Sermon statistics show how Christian congregations use online preaching to address scripture, worship, and public life. Pew Research Center studied 49,719 sermons shared online by 6,431 churches in a 2019 analysis, then examined 12,832 sermons, homilies, or services from 2,143 churches during fall 2020. The two studies measure different periods and questions, so their results should be read separately.
Contents
- How Pew built the sermon dataset
- How long sermons lasted
- How traditions differed in the 2019 sample
- Old and New Testament references
- Words and phrases across churches
- Sermons and public issues in fall 2020
How Pew built the sermon dataset
The 2019 findings come from a large collection of online sermons, but the research process involved several stages of identification, matching, scraping, and review. Pew’s initial Google Places search identified 478,699 institutions labeled as churches. Researchers matched 262,876 congregations from Google Places to the InfoGroup congregation database, which contained 539,778 congregations. Human coders manually matched 1,654 churches between Pew’s database and InfoGroup’s database.
Pew then deployed its website scraper to 38,630 American churches. To find sermon pages, researchers trained a machine-learning classifier on 906 links from 318 church websites. Of those links, 412 contained sermons and 494 did not. The classifier achieved 0.86 accuracy, 0.86 precision, and 0.83 recall.
The research team also checked the process for non-sermons. It randomly sampled 530 transcribed files for human review. The non-sermon filter achieved 90% accuracy, 92% recall, and 93% precision. These figures describe the study’s classification and filtering performance; they are not estimates of how many sermons exist across all congregations.
The resulting 2019 study analyzed 49,719 sermons shared online by 6,431 churches. Because the collection is based on sermons available online and the study used a defined research process, its figures describe the measured sample and should not automatically be treated as a census of every sermon delivered in the United States.
How long sermons lasted
The median sermon length in the 2019 study was 37 minutes. A median separates the sample into two equal halves: half of the measured sermons were shorter than the median and half were longer. The results varied substantially by major Christian tradition.
| Tradition | Median sermon length in 2019 |
|---|---|
| Catholic | 14 minutes |
| Mainline Protestant | 25 minutes |
| Evangelical Protestant | 39 minutes |
| Historically Black Protestant | 54 minutes |
Historically Black Protestant sermons had the longest reported median, at 54 minutes. Evangelical Protestant sermons had a 39-minute median, two minutes above the overall median. Mainline Protestant sermons had a 25-minute median, while Catholic sermons had a 14-minute median. These are medians for the 2019 study, not recommended service lengths or forecasts.
The sample also contained different numbers of sermons per congregation. The 2019 dataset represented 9.6 sermons per evangelical Protestant congregation, 8.1 sermons per historically Black Protestant congregation, 7.9 sermons per mainline Protestant congregation, and 6.4 sermons per Catholic congregation. Those averages help describe the structure of the collected dataset, while the median-length figures describe sermon duration.
How traditions differed in the 2019 sample
The raw sermon database was not evenly divided among the four traditions listed in the study. Evangelical Protestant sermons accounted for 59% of raw sermons in the 2019 database. Mainline Protestant sermons accounted for 16%. Catholic sermons accounted for 1%, and historically Black Protestant sermons also accounted for 1%.
These percentages are raw database shares, so they should not be read as the percentage distribution of all American congregations or all Christian sermons. They do, however, provide important context for reading comparisons. A result calculated from the complete raw collection may be influenced by how many sermons from each tradition entered the database.
The 2019 study also reported a congregation-size difference in Old Testament references. Smaller churches with 200 or fewer members cited Old Testament books in 6 percentage points more of their sermons on average than churches with more than 200 members. The reported difference is a comparison between those two church-size groups; the supplied figure does not specify the underlying percentages.
The measured sample therefore combines several kinds of information: the number of churches and sermons collected, the distribution of traditions in the raw database, sermon duration, and differences associated with congregation size. Keeping those measures separate prevents a database share from being mistaken for a theological or demographic prevalence rate.
Old and New Testament references
Biblical references were common in the 2019 sermon collection. At least one New Testament book was named in 90% of all sermons, while at least one Old Testament book was cited in 61% of all sermons. The figures refer to named books in the analyzed sermons, not to the total number of biblical references or the amount of time spent discussing scripture.
The pattern differed by tradition:
| Tradition | Old Testament book named | New Testament book named |
|---|---|---|
| Evangelical Protestant | 66% | 93% |
| Mainline Protestant | 43% | 88% |
| Catholic | 28% | 73% |
| Historically Black Protestant | 65% | 85% |
Evangelical sermons named an Old Testament book in 66% of sermons and a New Testament book in 93%. Mainline Protestant sermons named an Old Testament book in 43% and a New Testament book in 88%. Catholic homilies named an Old Testament book in 28% and a New Testament book in 73%. Historically Black Protestant sermons named an Old Testament book in 65% and a New Testament book in 85%.
The study supplied one additional combined measure: evangelical sermons named both an Old Testament book and a New Testament book in 62% of sermons. This combined percentage is a distinct measure from the separate 66% and 93% figures, because it counts sermons meeting both conditions.
The timing of the sermon also mattered in one reported comparison. During the Easter week studied, Old Testament mentions across all Christian groups fell to 49%, from 62% the week before. The figures describe the specific weeks examined in the 2019 study and should not be generalized to every Easter period.
Words and phrases across churches
The 2019 analysis also examined individual words and phrases appearing in sermons. “Jesus” appeared in sermons at 99% of congregations. The words “know,” “God,” and “Jesus” each appeared in sermons at 98% or more of churches across all four major traditions.
These results are congregation-level presence measures: they indicate how widely a word appeared across churches in the measured sample. They do not report how many times a word appeared in each sermon, nor do they measure the meaning or context of the surrounding discussion.
Some phrases were more concentrated by tradition. The phrase “hallelujah … come” appeared in 22% of historically Black Protestant churches. The supplied study figure refers to churches in that tradition, not to the share of all sermons and not to all Christian congregations.
Taken together, the word-level figures show both broad commonality and variation. Terms appearing at 98% or more of churches across traditions were widespread in the sample, while the reported phrase measure identifies a narrower pattern within historically Black Protestant churches. Neither type of result measures congregational belief, attendance, or the effectiveness of a sermon.
Sermons and public issues in fall 2020
Pew Research Center’s fall 2020 study analyzed 12,832 sermons, homilies, or services from 2,143 churches. It measured whether congregations heard at least one sermon touching on selected public issues during that period. In fall 2020, 83% of congregations heard at least one sermon touching on the COVID-19 pandemic, 67% heard at least one sermon mentioning the 2020 election, and 44% heard at least some discussion of racism in America.
The study also reported measures related to voting and political discussion. Twenty-three percent of all congregations heard at least one sermon encouraging voting. Forty-three percent heard at least one sermon discussing parties, issues, or candidates. These percentages describe congregations in the fall 2020 study, not the percentage of individual sermons with each theme.
Election references varied by tradition. The share of congregations hearing at least one sermon mentioning the election was 71% among evangelical Protestant congregations, 63% among mainline Protestant congregations, 63% among historically Black Protestant congregations, and 41% among Catholic congregations.
| Tradition | Congregations hearing at least one election sermon |
|---|---|
| Evangelical Protestant | 71% |
| Mainline Protestant | 63% |
| Historically Black Protestant | 63% |
| Catholic | 41% |
Among sermons that mentioned the election, 48% of evangelical Protestant sermons discussed specific issues, parties, or candidates. Among historically Black Protestant sermons that mentioned the election, 43% explicitly encouraged voting or assumed listeners would vote. These denominational figures use different denominators from the congregation-level election percentages, so they should not be directly substituted for one another.
Racism was mentioned in 32% of Catholic congregations in the fall 2020 study. For comparison, the all-congregation measure for at least some discussion of racism in America was 44%. The Catholic figure is a tradition-specific result, while the 44% figure covers the study’s overall congregation sample.
The weighted sample itself consisted of 52% evangelical Protestant congregations and 22% mainline Protestant congregations. Those composition figures provide context for the overall results. The fall 2020 study captured a defined period shaped by the COVID-19 pandemic, the 2020 election, and public discussion of racism; it does not provide a forecast for later sermon topics.
Reading sermon statistics responsibly
The two Pew Research Center studies use different samples and measurement periods. The 2019 analysis focuses on online sermons, sermon length, biblical references, words, phrases, and congregation characteristics. The fall 2020 analysis focuses on sermons, homilies, or services and whether congregations heard discussion of the pandemic, election, racism, voting, or political subjects.
The safest comparisons keep the unit of measurement visible. A percentage of sermons is not the same as a percentage of congregations. A raw database share is not the same as a population share. A median duration is not an average duration. Finally, a result from fall 2020 is a measurement of that period, not a prediction of future preaching patterns.