Chapter 9.3

Religious tradition and race

In 2024 white born-again Protestants voted 14.5% Democratic and Black Protestants 83.1%, and attendance relates to the vote by +9.5 points among the first and +0.8 points among the second, so the pooled attendance gap combines groups with different gaps.

Author: Vishal Singh
Affiliation: NYU Stern
Published: 3 October 2026

DraftFigures and numbers may change as the analysis is rebuilt.what is still being checked

The CES tradition coding is a proxy for denomination, and the sign of the attendance gap among white Protestants who are not born-again differs by survey; the last two sections describe the coding and its checks.

Among voters in the 2024 Cooperative Election Study (CES), white born-again Protestants voted 14.5% Democratic in the two-party presidential vote and Black Protestants voted 83.1%. Both groups attend services often: 57.8% of white born-again Protestant voters and 47.3% of Black Protestant voters attend weekly or more, against 26.7% of all voters. Attendance relates to the vote differently inside the two groups. Seldom attenders, who go a few times a year or never, voted more Democratic than frequent attenders by +9.5 (95% CI 6.6 to 12.4) points among white born-again Protestants and by +0.8 (95% CI −5.1 to 6.7) points among Black Protestants, an interval that includes zero. Among all voters the same gap, which Chapter 9.1, The god gap in America measures, was 27.7 (95% CI 26.0 to 29.4) points. This chapter asks what that single number combines.

Tradition and race are coded together. The CES has no denomination item, so white Protestants are split by whether they call themselves born-again or evangelical Christians: those who do are white born-again Protestants and those who do not are white Protestants who are not born-again. Black, Hispanic and Asian or other-race Protestants are kept whole, Catholics are split by race, and the remaining groups are Jewish, Mormon, Muslim, Orthodox Christian, and Buddhist or Hindu voters, voters who gave another religion, and the unaffiliated, who are atheists, agnostics and people with no religion in particular. White means non-Hispanic white. A gap in this chapter is the Democratic share among seldom attenders minus the share among frequent attenders, who attend weekly or more, so a positive gap means that seldom attenders voted more Democratic.The born-again rule stands in for denomination. A denominational scheme, RELTRAD [1], sorts Protestants into evangelical, mainline and Black Protestant traditions from the denomination they name. It can be coded in the General Social Survey (GSS), and the 2024 American National Election Studies (ANES) have their own major-group summary, so results from those surveys use the terms evangelical and mainline, and the last section checks the CES rule against both. Shares use the survey weights, and the CES weight is the one defined for all five elections. Every interval is a 95% confidence interval from the Kish effective sample size [2], and a cell needs at least 30 respondents. The CES is an opt-in online panel, so its intervals cover sampling error and not how respondents were recruited. The sample is 188,145 CES respondents who reported voting for the Democratic or the Republican candidate in the five presidential elections from 2008 to 2024, 42,008 of them in 2024.

Chapter 9.1, The god gap in America measures the overall gap and compares the United States with other countries, Chapter 9.2, The rise of the nones (planned) follows the unaffiliated, and Chapter 3.4, Hispanic, Black and Asian voters (planned) follows the vote of Black, Hispanic and Asian voters across traditions. This chapter takes the traditions one at a time and asks whether attendance relates to the vote in the same way inside each.

Two groups that attend often sit at opposite ends of the vote

Figure 1 orders the groups by their Democratic share in 2024. White born-again Protestants were at 14.5% (95% CI 13.2% to 15.7%), white Catholics at 37.1% (95% CI 35.4% to 38.9%), white Protestants who are not born-again at 43.0% (95% CI 41.0% to 45.0%), Hispanic Catholics at 57.0% (95% CI 52.5% to 61.5%), the unaffiliated at 70.9% (95% CI 69.7% to 72.2%) and Black Protestants at 83.1% (95% CI 80.3% to 86.0%). Party identification gives the same order, from 13.7% Democratic among white born-again Protestants to 86.3% among Black Protestants (the third view). Pooled over the five elections the vote shares run from 18.1% (95% CI 17.6% to 18.7%) among white born-again Protestants to 91.4% (95% CI 90.4% to 92.3%) among Black Protestants (the second view).

Figure 1

In 2024 the Democratic share of the two-party vote ran from 14.5% among white born-again Protestants to 83.1% among Black Protestants

Democratic share of the two-party presidential vote by religious tradition and race, reported voters, percent, 2024 (buttons: five elections pooled, and party identification); 95% intervals

Protestant
Catholic
Other religion
Unaffiliated

Show data

Data for In 2024 the Democratic share of the two-party vote ran from 14.5% among white born-again Protestants to 83.1% among Black Protestants
View	Group	Democratic share (%)	Lower	Upper	Interval width (points)	Respondents	Group share of the sample (%)
vote2024	Black Protestants	83.1	80.3	86.0	5.7	1,889	4.7
vote2024	Black Catholics	76.0	67.2	84.8	17.7	260	0.6
vote2024	Buddhist or Hindu	71.6	64.8	78.5	13.7	431	1.1
vote2024	Unaffiliated	70.9	69.7	72.2	2.4	14,311	31.0
vote2024	Jewish	68.0	64.0	72.1	8.1	1,219	2.7
vote2024	Muslim	65.9	52.7	79.0	26.3	139	0.5
vote2024	Hispanic Catholics	57.0	52.5	61.5	9.1	1,351	4.0
vote2024	Asian or other-race Catholics	45.4	37.0	53.7	16.7	388	1.2
vote2024	Another religion or something else	43.9	40.8	46.9	6.2	3,082	7.7
vote2024	White Protestants who are not born-again	43.0	41.0	45.0	4.0	4,979	9.9
vote2024	White Catholics	37.1	35.4	38.9	3.5	5,974	12.7
vote2024	Hispanic Protestants	33.1	27.3	38.9	11.7	646	2.1
vote2024	Mormon	31.1	24.8	37.4	12.5	438	1.1
vote2024	Orthodox Christian	30.9	19.7	42.0	22.3	197	0.5
vote2024	Asian or other-race Protestants	30.4	25.0	35.9	10.9	699	2.1
vote2024	White born-again Protestants	14.5	13.2	15.7	2.5	5,987	18.1
votepool	Black Protestants	91.4	90.4	92.3	1.8	8,852	5.0
votepool	Black Catholics	88.4	85.5	91.3	5.8	1,105	0.6
votepool	Muslim	81.4	76.7	86.2	9.5	600	0.4
votepool	Buddhist or Hindu	78.9	76.2	81.6	5.4	1,827	0.9
votepool	Unaffiliated	72.2	71.6	72.8	1.1	53,647	26.8
votepool	Jewish	69.8	68.1	71.4	3.3	5,325	2.4
votepool	Hispanic Catholics	67.0	65.1	68.8	3.8	6,487	3.8
votepool	Another religion or something else	55.4	54.2	56.7	2.5	13,375	8.5
votepool	Asian or other-race Catholics	49.6	46.0	53.1	7.1	1,706	0.9
votepool	White Protestants who are not born-again	46.4	45.6	47.2	1.6	25,359	12.0
votepool	Hispanic Protestants	42.9	40.0	45.9	5.9	2,923	1.6
votepool	Orthodox Christian	42.1	37.4	46.9	9.5	879	0.5
votepool	White Catholics	41.4	40.6	42.1	1.5	29,810	14.8
votepool	Asian or other-race Protestants	34.8	32.4	37.3	4.9	3,172	1.7
votepool	Mormon	24.5	22.2	26.7	4.5	2,527	1.5
votepool	White born-again Protestants	18.1	17.6	18.7	1.1	30,356	18.4
party2024	Black Protestants	86.3	83.9	88.6	4.7	3,072	5.3
party2024	Black Catholics	80.7	73.3	88.1	14.9	427	0.8
party2024	Buddhist or Hindu	72.8	66.8	78.8	12.0	604	1.3
party2024	Unaffiliated	71.6	70.4	72.7	2.4	17,677	32.0
party2024	Jewish	68.9	64.8	73.0	8.2	1,353	2.5
party2024	Muslim	64.7	54.8	74.6	19.8	300	0.6
party2024	Hispanic Catholics	63.4	59.6	67.3	7.8	2,325	4.8
party2024	Asian or other-race Catholics	48.8	40.7	56.9	16.2	482	1.2
party2024	Another religion or something else	46.1	43.1	49.1	6.0	3,770	7.6
party2024	White Protestants who are not born-again	42.8	40.8	44.9	4.1	5,297	9.2
party2024	White Catholics	37.6	35.7	39.4	3.7	6,296	11.6
party2024	Hispanic Protestants	37.5	32.0	43.0	11.0	1,109	2.5
party2024	Orthodox Christian	37.3	25.4	49.1	23.7	257	0.6
party2024	Asian or other-race Protestants	34.0	28.2	39.7	11.5	825	1.9
party2024	Mormon	30.3	24.3	36.3	12.1	523	1.1
party2024	White born-again Protestants	13.7	12.5	15.0	2.5	6,484	16.8

The groups run from 14.5% to 83.1% Democratic, and inside each large Christian family the vote differs by race: Black and white Protestants, Hispanic and white Catholics. Each dot is a weighted share with a 95% interval, and the colour marks the tradition family. White means non-Hispanic white.Note: White born-again Protestants and white Protestants who are not born-again are the white Protestants who do and do not call themselves born-again or evangelical, because the CES file has no denomination; the last figure compares this coding with denominational codings. Black, Hispanic and Asian or other-race Protestants and the Catholic groups are not split further.Note: The interval width is in the data table. A cell is drawn only when it has at least 30 respondents and an effective sample of at least 30. Intervals are 95% intervals from the Kish effective sample size and treat the weighted sample as a simple random sample. The CES is an opt-in online panel, so they cover sampling error and not how respondents were recruited.Source: Cooperative Election Study, presidential years 2008, 2012, 2016, 2020 and 2024; weighted (weight_cumulative)

Race separates voters inside each large Christian family. Among Catholics, Hispanic Catholics voted 19.9 (95% CI 15.0 to 24.7) points more Democratic than white Catholics, and the 2024 estimate for Black Catholics, 76.0% (95% CI 67.2% to 84.8%), rests on a small sample. Among Protestants, Hispanic Protestants voted 33.1% (95% CI 27.3% to 38.9%) Democratic, above white born-again Protestants and below Black Protestants. Hispanic voters divide by tradition: Hispanic Catholics voted 23.9 (95% CI 16.5 to 31.3) points more Democratic than Hispanic Protestants. Among white Protestants, those who are not born-again voted 28.5 (95% CI 26.1 to 30.8) points more Democratic than those who are.

Among the smaller traditions, Jewish voters were 68.0% (95% CI 64.0% to 72.1%) Democratic and Mormon voters 31.1% (95% CI 24.8% to 37.4%), both inside the range that white born-again Protestants and Black Protestants set. Muslim voters were 65.9% (95% CI 52.7% to 79.0%), and that interval is the widest in the figure.

Attendance does not order the groups. Figure 2 plots the share of voters in each group who attend weekly or more against the Democratic share. Black Protestant voters attend weekly or more at 47.3% (95% CI 43.6% to 51.1%) and say religion is very important in their life at 72.3%, and 70.6% call themselves born-again or evangelical, the answer that defines white born-again Protestants here. For white born-again Protestants the first two shares are 57.8% (95% CI 56.0% to 59.5%) and 79.8%. Among all voters they are 26.7% and 39.0%. Hispanic Protestant voters (55.6% weekly) and Mormon voters (67.7%) also attend often and voted 33.1% and 31.1% Democratic, while the unaffiliated, of whom 2.5% attend weekly, voted 70.9%. The groups with the highest Democratic shares include one with a high attendance rate and one with a low one.

Figure 2

White born-again and Black Protestant voters attend weekly or more at 57.8% and 47.3%, against 26.7% of all voters, and voted 14.5% and 83.1% Democratic

Share of reported voters who attend services weekly or more (horizontal) against the Democratic share of the two-party presidential vote (vertical), by tradition and race, 2024; dot area is the group's share of voters

Protestant
Catholic
Other religion
Unaffiliated
Dot area: Share of voters (%)3.111.031.0

Show data

Data for White born-again and Black Protestant voters attend weekly or more at 57.8% and 47.3%, against 26.7% of all voters, and voted 14.5% and 83.1% Democratic
Group	Weekly or more (%)	Lower	Upper	Democratic share (%)	Lower	Upper	Share of voters (%)	Respondents
White born-again Protestants	57.8	56.0	59.5	14.5	13.2	15.7	18.1	5,987
White Protestants who are not born-again	18.4	16.9	20.0	43.0	41.0	45.0	9.9	4,979
Black Protestants	47.3	43.6	51.1	83.1	80.3	86.0	4.7	1,889
Hispanic Protestants	55.6	49.4	61.8	33.1	27.3	38.9	2.1	646
Asian or other-race Protestants	45.8	39.8	51.7	30.4	25.0	35.9	2.1	699
White Catholics	28.1	26.4	29.7	37.1	35.4	38.9	12.7	5,974
Hispanic Catholics	29.4	25.2	33.6	57.0	52.5	61.5	4.0	1,351
Black Catholics	42.2	32.0	52.5	76.0	67.2	84.8	0.6	260
Asian or other-race Catholics	36.4	28.3	44.5	45.4	37.0	53.7	1.2	388
Jewish	13.4	10.4	16.3	68.0	64.0	72.1	2.7	1,219
Mormon	67.7	61.4	74.0	31.1	24.8	37.4	1.1	438
Muslim	35.4	22.2	48.7	65.9	52.7	79.0	0.5	139
Orthodox Christian	41.9	29.8	54.0	30.9	19.7	42.0	0.5	197
Buddhist or Hindu	8.8	4.5	13.0	71.6	64.8	78.5	1.1	431
Another religion or something else	29.6	26.7	32.4	43.9	40.8	46.9	7.7	3,082
Unaffiliated	2.5	2.1	3.0	70.9	69.7	72.2	31.0	14,311

Groups with high attendance sit far apart on the vote: Mormon voters, white born-again Protestants, Hispanic Protestants and Black Protestants have the four highest weekly attendance point estimates and range from 14.5% to 83.1% Democratic. Each dot is one group in the 2024 CES, and the data table gives the intervals for both axes.Note: Voters who gave no attendance answer are left out of the attendance shares. A cell is drawn only when it has at least 30 respondents and an effective sample of at least 30.Note: Intervals are 95% intervals from the Kish effective sample size and treat the weighted sample as a simple random sample. The CES is an opt-in online panel, so they cover sampling error and not how respondents were recruited.Source: Cooperative Election Study, 2024; weighted (weight_cumulative)

Two groups with weekly attendance of 57.8% and 47.3%, against 26.7% among all voters, voted 68.6 (95% CI 65.6 to 71.7) points apart. The data cannot say why. Accounts of the Black church describe an institution of community and political life as well as of worship [3], but the survey items here measure attendance, importance and identification and not what congregations do.

The five largest Christian groups moved toward the Republican candidate by different amounts

Figure 3 compares each later election with 2008. Over the full period the Democratic share changed by −13.2 (95% CI −19.5 to −6.9) points among Hispanic Catholics, −10.3 (95% CI −14.1 to −6.5) among Black Protestants, −7.9 (95% CI −10.7 to −5.1) among white Protestants who are not born-again, −6.5 (95% CI −8.9 to −4.0) among white Catholics and −6.2 (95% CI −8.1 to −4.4) among white born-again Protestants, and each of those intervals lies below zero. The five groups and the unaffiliated, who changed by −5.6 (95% CI −7.6 to −3.7) points, kept the same order in 2008 and 2024, from Black Protestants at 93.4% and 83.1% to white born-again Protestants at 20.7% and 14.5%.

Figure 3

From 2008 to 2024 the Democratic share changed by −13.2 points among Hispanic Catholics and −10.3 among Black Protestants, against −6.2 among white born-again Protestants and −6.5 among white Catholics

Democratic share of the two-party presidential vote by tradition and race, reported voters, percent, 2008 and a later election (the buttons switch the later election); 95% intervals

2008
2024

Show data

Data for From 2008 to 2024 the Democratic share changed by −13.2 points among Hispanic Catholics and −10.3 among Black Protestants, against −6.2 among white born-again Protestants and −6.5 among white Catholics
Later election	Group	Share 2008 (%)	Lower 2008	Upper 2008	Respondents 2008	Share later (%)	Lower later	Upper later	Respondents later	Change (points)	Lower change	Upper change
2024	White born-again Protestants	20.7	19.3	22.1	4,399	14.5	13.2	15.7	5,987	−6.2	−8.1	−4.4
2024	White Protestants who are not born-again	50.9	48.9	52.8	3,353	43.0	41.0	45.0	4,979	−7.9	−10.7	−5.1
2024	Black Protestants	93.4	90.9	95.9	456	83.1	80.3	86.0	1,889	−10.3	−14.1	−6.5
2024	Hispanic Protestants	36.4	28.7	44.1	176	33.1	27.3	38.9	646	−3.3	−13.0	6.4
2024	Asian or other-race Protestants	29.3	23.7	34.9	340	30.4	25.0	35.9	699	1.2	−6.7	9.0
2024	White Catholics	43.6	41.9	45.3	4,264	37.1	35.4	38.9	5,974	−6.5	−8.9	−4.0
2024	Hispanic Catholics	70.2	65.8	74.6	599	57.0	52.5	61.5	1,351	−13.2	−19.5	−6.9
2024	Black Catholics	94.5	89.3	99.6	94	76.0	67.2	84.8	260	−18.5	−28.7	−8.2
2024	Asian or other-race Catholics	48.8	40.3	57.3	219	45.4	37.0	53.7	388	−3.4	−15.4	8.5
2024	Jewish	71.3	67.1	75.6	552	68.0	64.0	72.1	1,219	−3.3	−9.2	2.5
2024	Mormon	24.6	20.1	29.1	465	31.1	24.8	37.4	438	6.5	−1.2	14.2
2024	Muslim	96.0	89.4	100.0	50	65.9	52.7	79.0	139	−30.1	−44.8	−15.4
2024	Orthodox Christian	51.8	40.4	63.2	91	30.9	19.7	42.0	197	−20.9	−36.9	−4.9
2024	Buddhist or Hindu	91.1	85.9	96.3	150	71.6	64.8	78.5	431	−19.5	−28.1	−10.9
2024	Another religion or something else	58.4	56.4	60.4	3,778	43.9	40.8	46.9	3,082	−14.5	−18.2	−10.9
2024	Unaffiliated	76.5	75.0	78.1	4,544	70.9	69.7	72.2	14,311	−5.6	−7.6	−3.7
2020	White born-again Protestants	20.7	19.3	22.1	4,399	15.7	14.7	16.7	6,263	−5.0	−6.7	−3.3
2020	White Protestants who are not born-again	50.9	48.9	52.8	3,353	46.7	45.1	48.2	5,132	−4.2	−6.7	−1.7
2020	Black Protestants	93.4	90.9	95.9	456	91.3	89.5	93.0	2,191	−2.1	−5.2	0.9
2020	Hispanic Protestants	36.4	28.7	44.1	176	51.5	45.6	57.5	778	15.1	5.4	24.9
2020	Asian or other-race Protestants	29.3	23.7	34.9	340	35.4	30.4	40.4	713	6.1	−1.4	13.6
2020	White Catholics	43.6	41.9	45.3	4,264	40.4	39.0	41.8	6,265	−3.2	−5.4	−1.0
2020	Hispanic Catholics	70.2	65.8	74.6	599	66.2	62.2	70.2	1,621	−4.0	−9.9	1.9
2020	Black Catholics	94.5	89.3	99.6	94	88.6	83.1	94.0	262	−5.9	−13.4	1.7
2020	Asian or other-race Catholics	48.8	40.3	57.3	219	54.5	47.6	61.4	404	5.7	−5.2	16.7
2020	Jewish	71.3	67.1	75.6	552	71.5	68.9	74.2	1,323	0.2	−4.8	5.2
2020	Mormon	24.6	20.1	29.1	465	28.5	23.6	33.4	496	3.9	−2.8	10.5
2020	Muslim	96.0	89.4	100.0	50	84.8	76.0	93.6	179	−11.1	−22.1	−0.2
2020	Orthodox Christian	51.8	40.4	63.2	91	40.9	30.0	51.7	202	−10.9	−26.7	4.8
2020	Buddhist or Hindu	91.1	85.9	96.3	150	74.3	69.2	79.4	495	−16.8	−24.1	−9.5
2020	Another religion or something else	58.4	56.4	60.4	3,778	51.7	49.1	54.2	2,716	−6.7	−9.9	−3.5
2020	Unaffiliated	76.5	75.0	78.1	4,544	73.1	72.1	74.0	14,815	−3.4	−5.2	−1.7
2016	White born-again Protestants	20.7	19.3	22.1	4,399	17.5	16.3	18.7	6,313	−3.2	−5.1	−1.4
2016	White Protestants who are not born-again	50.9	48.9	52.8	3,353	43.4	41.8	45.1	5,568	−7.5	−10.0	−4.9
2016	Black Protestants	93.4	90.9	95.9	456	92.9	91.2	94.6	2,337	−0.5	−3.5	2.6
2016	Hispanic Protestants	36.4	28.7	44.1	176	46.1	39.9	52.2	774	9.7	−0.2	19.6
2016	Asian or other-race Protestants	29.3	23.7	34.9	340	39.9	35.1	44.6	714	10.6	3.3	18.0
2016	White Catholics	43.6	41.9	45.3	4,264	41.4	40.0	42.9	6,737	−2.2	−4.4	0.1
2016	Hispanic Catholics	70.2	65.8	74.6	599	73.2	69.3	77.1	1,725	3.0	−2.8	8.8
2016	Black Catholics	94.5	89.3	99.6	94	88.3	82.3	94.3	294	−6.1	−14.1	1.8
2016	Asian or other-race Catholics	48.8	40.3	57.3	219	51.6	45.1	58.1	373	2.8	−7.9	13.5
2016	Jewish	71.3	67.1	75.6	552	71.8	68.8	74.8	1,200	0.5	−4.7	5.7
2016	Mormon	24.6	20.1	29.1	465	31.3	25.6	37.0	447	6.7	−0.6	13.9
2016	Muslim	96.0	89.4	100.0	50	86.3	79.6	92.9	163	−9.7	−19.0	−0.4
2016	Orthodox Christian	51.8	40.4	63.2	91	40.5	32.3	48.6	222	−11.3	−25.3	2.7
2016	Buddhist or Hindu	91.1	85.9	96.3	150	77.6	72.5	82.8	435	−13.5	−20.8	−6.2
2016	Another religion or something else	58.4	56.4	60.4	3,778	58.1	55.1	61.0	2,022	−0.3	−3.9	3.3
2016	Unaffiliated	76.5	75.0	78.1	4,544	68.2	67.1	69.4	11,528	−8.3	−10.2	−6.4
2012	White born-again Protestants	20.7	19.3	22.1	4,399	22.2	20.8	23.6	7,394	1.5	−0.4	3.5
2012	White Protestants who are not born-again	50.9	48.9	52.8	3,353	47.8	46.1	49.5	6,327	−3.1	−5.7	−0.5
2012	Black Protestants	93.4	90.9	95.9	456	95.9	94.6	97.3	1,979	2.5	−0.3	5.4
2012	Hispanic Protestants	36.4	28.7	44.1	176	46.7	39.8	53.7	549	10.3	−0.1	20.7
2012	Asian or other-race Protestants	29.3	23.7	34.9	340	37.8	32.0	43.5	706	8.5	0.5	16.5
2012	White Catholics	43.6	41.9	45.3	4,264	43.3	41.7	45.0	6,570	−0.3	−2.6	2.1
2012	Hispanic Catholics	70.2	65.8	74.6	599	69.6	65.5	73.8	1,191	−0.6	−6.5	5.4
2012	Black Catholics	94.5	89.3	99.6	94	96.1	92.1	100.0	195	1.6	−4.9	8.2
2012	Asian or other-race Catholics	48.8	40.3	57.3	219	49.1	40.7	57.5	322	0.3	−11.7	12.3
2012	Jewish	71.3	67.1	75.6	552	66.7	62.5	70.9	1,031	−4.7	−10.6	1.3
2012	Mormon	24.6	20.1	29.1	465	15.2	11.2	19.1	681	−9.4	−15.4	−3.5
2012	Muslim	96.0	89.4	100.0	50	81.1	67.7	94.6	69	−14.8	−29.8	0.1
2012	Orthodox Christian	51.8	40.4	63.2	91	50.2	39.0	61.4	167	−1.6	−17.6	14.4
2012	Buddhist or Hindu	91.1	85.9	96.3	150	86.8	81.1	92.6	316	−4.3	−12.1	3.4
2012	Another religion or something else	58.4	56.4	60.4	3,778	64.3	60.8	67.7	1,777	5.9	1.9	9.9
2012	Unaffiliated	76.5	75.0	78.1	4,544	73.6	72.2	75.1	8,449	−2.9	−5.0	−0.8

Each of the five largest Christian groups moved toward the Republican candidate over the period, and the point estimates of the moves are largest among Hispanic Catholics and Black Protestants; the text tests the differences between the moves. The ring is 2008 and the dot the later election; a group is drawn when it meets the sample floor in both years, and the table gives the change with its interval.Note: In the default view, 2008 and 2024, Mormon voters are the one group whose point estimate rose by more than a few points, and the interval of the change includes zero; in the other views other groups rise. A cell is drawn only when it has at least 30 respondents and an effective sample of at least 30.Note: Intervals are 95% intervals from the Kish effective sample size and treat the weighted sample as a simple random sample. The CES is an opt-in online panel, so they cover sampling error and not how respondents were recruited.Source: Cooperative Election Study, presidential years 2008, 2012, 2016, 2020 and 2024; weighted (weight_cumulative)

The two largest falls are those of Hispanic Catholics and Black Protestants, and the differences between the changes can be tested. The change among Hispanic Catholics differs from the change among white born-again Protestants by −7.0 (95% CI −13.5 to −0.4) points and from the change among the unaffiliated by −7.6 (95% CI −14.2 to −1.0), and the change among Black Protestants differs from the change among the unaffiliated by −4.7 (95% CI −8.9 to −0.4). Hispanic Catholics differ from white Catholics by −6.7 (95% CI −13.5 to 0.0), an interval that reaches zero, and every other difference among the six changes has an interval that includes zero. The groups that moved most are therefore separable from some of the others and not from all of them.

The order does not hold for every group. Muslim voters, Black Catholic voters and Buddhist or Hindu voters were among the most Democratic in 2008 and changed by −30.1 (95% CI −44.8 to −15.4), −18.5 (95% CI −28.7 to −8.2) and −19.5 (95% CI −28.1 to −10.9) points, so the order of all the groups differs between 2008 and 2024. Hispanic Protestants, Jewish voters and Mormon voters show no change that the sample can separate from zero: −3.3 (95% CI −13.0 to 6.4), −3.3 (95% CI −9.2 to 2.5) and +6.5 (95% CI −1.2 to 14.2) points, the last being the only point estimate that rose by more than a few points. Chapter 3.4, Hispanic, Black and Asian voters (planned) follows Black and Hispanic voters as a whole, and the figure here describes the same voters split by tradition.

Attendance relates to the vote in different directions inside each group

Figure 4 splits each group by attendance. One sign comes from the CES coding and is not reproduced in the other surveys: the gap among white Protestants who are not born-again, taken up below and in the last section. Pooled over the five elections, seldom attenders voted more Democratic than frequent attenders by +10.1 (95% CI 8.7 to 11.4) points among white born-again Protestants, +13.7 (95% CI 12.2 to 15.3) among white Catholics, +11.9 (95% CI 7.5 to 16.3) among Hispanic Catholics and +13.8 (95% CI 7.5 to 20.0) among Hispanic Protestants. The white Catholic gap exceeds the white born-again gap by 3.6 (95% CI 1.6 to 5.7) points, and the two Hispanic intervals are wide enough to include both. In 2024 alone the four gaps were +9.5 (95% CI 6.6 to 12.4), +12.0 (95% CI 8.1 to 15.9), +15.3 (95% CI 4.9 to 25.7) and +10.6 (95% CI −2.0 to 23.2).

Figure 4

Inside tradition-by-race groups the vote differs by attendance in different directions: seldom minus frequent attenders in the Democratic share was +13.7 points among white Catholics, +10.1 among white born-again Protestants, +1.9 among Black Protestants and −5.4 among white Protestants who are not born-again, a sign that the GSS and ANES do not reproduce

Democratic share of the two-party presidential vote among frequent attenders (weekly or more) and seldom attenders (a few times a year or never), by tradition and race, reported voters, percent, 2008 to 2024 pooled (buttons: single years and party identification); 95% intervals

Weekly or more
A few times a year or never

Show data

Data for Inside tradition-by-race groups the vote differs by attendance in different directions: seldom minus frequent attenders in the Democratic share was +13.7 points among white Catholics, +10.1 among white born-again Protestants, +1.9 among Black Protestants and −5.4 among white Protestants who are not born-again, a sign that the GSS and ANES do not reproduce
View	Group	Frequent attenders (%)	Lower	Upper	Respondents	Seldom attenders (%)	Lower	Upper	Respondents	Seldom minus frequent (points)	Lower	Upper	Interval width (points)
votepool	Jewish	49.0	43.3	54.7	581	72.8	71.0	74.6	4,178	23.8	17.9	29.8	11.9
votepool	Unaffiliated	55.0	51.1	59.0	1,372	73.1	72.5	73.7	50,916	18.1	14.1	22.1	8.0
votepool	Mormon	19.7	17.3	22.1	1,830	37.2	31.2	43.2	545	17.5	11.1	23.9	12.8
votepool	Another religion or something else	44.6	42.4	46.8	3,889	61.4	59.8	62.9	8,245	16.8	14.1	19.5	5.4
votepool	Asian or other-race Protestants	28.0	24.6	31.3	1,480	42.6	38.6	46.7	1,332	14.7	9.4	19.9	10.5
votepool	Orthodox Christian	33.6	24.0	43.1	258	47.4	41.4	53.4	498	13.8	2.6	25.1	22.6
votepool	Hispanic Protestants	37.0	32.9	41.1	1,455	50.8	46.1	55.5	1,085	13.8	7.5	20.0	12.5
votepool	White Catholics	32.5	31.3	33.7	9,448	46.3	45.3	47.3	17,363	13.7	12.2	15.3	3.1
votepool	Asian or other-race Catholics	41.8	36.2	47.4	661	55.4	50.6	60.3	862	13.6	6.2	21.0	14.8
votepool	Hispanic Catholics	59.0	55.3	62.7	1,782	70.9	68.5	73.3	3,877	11.9	7.5	16.3	8.8
votepool	White born-again Protestants	14.2	13.6	14.9	17,922	24.3	23.2	25.5	9,263	10.1	8.7	11.4	2.7
votepool	Black Protestants	90.7	89.3	92.0	4,310	92.6	91.2	94.0	3,398	1.9	0.0	3.8	3.9
votepool	Black Catholics	88.0	83.0	93.0	401	89.6	85.7	93.5	550	1.6	−4.7	8.0	12.7
votepool	Buddhist or Hindu	79.6	72.4	86.7	217	78.1	75.0	81.2	1,442	−1.5	−9.3	6.3	15.6
votepool	Muslim	81.8	74.5	89.2	230	79.3	72.1	86.5	296	−2.5	−12.8	7.7	20.5
votepool	White Protestants who are not born-again	50.5	48.7	52.4	4,868	45.1	44.1	46.0	17,790	−5.4	−7.5	−3.4	4.1
vote2024	Unaffiliated	42.0	30.9	53.1	271	72.1	70.9	73.4	13,762	30.1	19.0	41.3	22.3
vote2024	Jewish	41.9	28.4	55.4	134	71.7	67.4	76.0	956	29.8	15.6	44.0	28.3
vote2024	Another religion or something else	32.1	26.8	37.4	804	50.0	46.1	53.9	2,022	17.9	11.4	24.5	13.1
vote2024	Hispanic Catholics	47.8	39.1	56.6	349	63.2	57.5	68.8	844	15.3	4.9	25.7	20.8
vote2024	Asian or other-race Protestants	23.0	15.2	30.8	287	36.4	28.2	44.5	346	13.4	2.1	24.6	22.5
vote2024	White Catholics	29.0	25.8	32.1	1,572	41.0	38.7	43.3	3,876	12.0	8.1	15.9	7.8
vote2024	Mormon	27.5	20.9	34.1	285	38.6	23.3	54.0	131	11.1	−5.6	27.8	33.4
vote2024	Hispanic Protestants	29.2	21.1	37.2	316	39.8	30.1	49.5	254	10.6	−2.0	23.2	25.2
vote2024	White born-again Protestants	10.7	9.3	12.2	3,267	20.2	17.8	22.7	2,113	9.5	6.6	12.4	5.7
vote2024	Asian or other-race Catholics	42.5	28.7	56.4	134	49.4	38.6	60.1	217	6.8	−10.7	24.3	35.0
vote2024	Black Protestants	83.4	79.1	87.6	839	84.2	80.1	88.3	845	0.8	−5.1	6.7	11.8
vote2024	Black Catholics	79.4	65.2	93.6	91	77.2	65.0	89.3	137	−2.3	−20.9	16.4	37.3
vote2024	White Protestants who are not born-again	46.6	42.3	50.9	904	42.5	40.2	44.8	3,588	−4.1	−9.0	0.8	9.8
vote2008	Asian or other-race Catholics	34.5	21.3	47.6	92	61.3	49.1	73.5	101	26.8	8.9	44.7	35.8
vote2008	Jewish	49.4	32.0	66.8	58	74.2	69.6	78.7	422	24.7	6.7	42.7	36.0
vote2008	Asian or other-race Protestants	21.1	13.9	28.3	181	44.0	33.9	54.1	114	22.9	10.5	35.3	24.8
vote2008	White Catholics	33.1	30.4	35.8	1,507	50.1	47.7	52.5	2,217	17.0	13.4	20.6	7.3
vote2008	Another religion or something else	49.6	46.6	52.6	1,664	66.2	63.3	69.1	1,676	16.6	12.4	20.8	8.4
vote2008	Hispanic Catholics	59.4	51.1	67.7	184	75.9	70.4	81.4	335	16.5	6.5	26.5	19.9
vote2008	Mormon	19.9	15.2	24.5	366	36.4	23.2	49.5	69	16.5	2.6	30.4	27.9
vote2008	White born-again Protestants	15.6	14.1	17.1	2,823	30.7	27.5	33.9	1,082	15.1	11.5	18.6	7.1
vote2008	Hispanic Protestants	33.1	22.5	43.8	87	42.2	27.4	57.0	56	9.1	−9.1	27.3	36.4
vote2008	Black Protestants	91.6	87.8	95.5	229	96.5	93.3	99.7	156	4.9	−0.1	9.9	10.1
vote2008	Unaffiliated	76.2	64.3	88.0	80	76.6	75.0	78.1	4,351	0.4	−11.5	12.3	23.8
vote2008	White Protestants who are not born-again	54.2	49.8	58.6	648	49.8	47.4	52.2	2,298	−4.4	−9.4	0.6	10.0
partypool	Mormon	17.2	15.0	19.3	2,269	41.3	35.8	46.8	684	24.1	18.3	30.0	11.7
partypool	Jewish	51.3	45.9	56.7	734	74.2	72.4	75.9	4,510	22.8	17.2	28.5	11.4
partypool	Asian or other-race Catholics	44.1	38.8	49.3	752	60.3	55.7	64.8	1,043	16.2	9.3	23.1	13.8
partypool	Asian or other-race Protestants	31.5	28.1	34.9	1,651	47.6	43.6	51.6	1,475	16.1	10.8	21.3	10.5
partypool	Another religion or something else	49.0	46.9	51.0	4,733	64.1	62.7	65.5	10,365	15.1	12.6	17.6	5.0
partypool	Hispanic Protestants	40.8	37.1	44.4	2,136	54.8	50.6	59.1	1,641	14.1	8.5	19.6	11.2
partypool	White Catholics	36.4	35.2	37.6	10,228	49.3	48.4	50.3	19,111	13.0	11.4	14.5	3.1
partypool	Unaffiliated	61.5	58.3	64.8	2,102	74.3	73.8	74.9	61,418	12.8	9.5	16.1	6.6
partypool	Orthodox Christian	38.8	31.1	46.4	358	51.2	45.3	57.1	634	12.4	2.8	22.1	19.3
partypool	White born-again Protestants	15.5	14.8	16.2	19,881	26.7	25.5	27.8	10,637	11.2	9.9	12.5	2.7
partypool	Hispanic Catholics	64.8	61.7	67.9	2,670	75.0	73.0	76.9	5,989	10.2	6.5	13.8	7.3
partypool	Muslim	77.6	71.6	83.5	508	83.2	78.3	88.1	537	5.6	−2.1	13.3	15.5
partypool	Buddhist or Hindu	74.2	67.7	80.7	329	78.1	75.3	81.0	1,949	3.9	−3.2	11.1	14.3
partypool	Black Protestants	90.7	89.6	91.9	6,552	93.3	92.2	94.4	5,409	2.6	1.0	4.2	3.2
partypool	Black Catholics	87.0	82.7	91.2	683	89.2	85.8	92.7	867	2.3	−3.2	7.7	10.9
partypool	White Protestants who are not born-again	49.6	47.7	51.4	5,161	45.9	44.9	46.8	19,334	−3.7	−5.8	−1.7	4.1

The gap between the ring (frequent attenders) and the dot (seldom attenders) is positive for most groups, small among Black Protestants and negative in the CES among white Protestants who are not born-again, a sign that the GSS and ANES do not reproduce (last figure). Each row is one group with its frequent and seldom shares and a 95% interval on each, and the last column of the table is seldom minus frequent, in points.Note: Monthly attenders are in neither group. Frequent is the top two answers of the CES six-point item, once a week and more than once a week. A group is drawn when both attendance cells meet the sample floor, so fewer groups appear in a single year. A cell is drawn only when it has at least 30 respondents and an effective sample of at least 30.Note: Intervals are 95% intervals from the Kish effective sample size and treat the weighted sample as a simple random sample. The CES is an opt-in online panel, so they cover sampling error and not how respondents were recruited.Source: Cooperative Election Study, presidential years 2008, 2012, 2016, 2020 and 2024; weighted (weight_cumulative)

Among Black Protestants the pooled gap is small, +1.9 (95% CI 0.0 to 3.8) points, with frequent attenders at 90.7% and seldom attenders at 92.6% Democratic, and for party identification it is +2.6 (95% CI 1.0 to 4.2) points. The vote interval lies below the intervals of the four groups above, so Black Protestants who attend often and seldom voted within a few points of one another, while the same two attendance groups among white born-again Protestants voted +10.1 points apart. In 2024 alone the Black Protestant gap was +0.8 (95% CI −5.1 to 6.7) points. In 2008 the gaps were +4.9 (95% CI −0.1 to 9.9) for Black Protestants, +15.1 (95% CI 11.5 to 18.6) for white born-again Protestants and −4.4 (95% CI −9.4 to 0.6) for white Protestants who are not born-again, the same order as in the pooled estimates.

Among white Protestants who are not born-again the direction reverses in the CES, and this is the sign that the GSS and the ANES do not reproduce: the GSS gives +2.8 (95% CI −3.8 to 9.4) points for the RELTRAD mainline group and +5.7 (95% CI −0.5 to 12.0) under the born-again rule, and the ANES gives +11.6 (95% CI −3.4 to 26.6) (the last section). In the CES, frequent attenders voted 50.5% Democratic and seldom attenders 45.1%, a gap of −5.4 (95% CI −7.5 to −3.4) points, and the point estimate is below zero in each of the five elections. The gap was −8.1 (95% CI −12.2 to −4.0) points in 2020 and −4.1 (95% CI −9.0 to 0.8) in 2024. Party identification gives the same signs in the CES: +11.2 (95% CI 9.9 to 12.5) among white born-again Protestants, +2.6 (95% CI 1.0 to 4.2) among Black Protestants and −3.7 (95% CI −5.8 to −1.7) among white Protestants who are not born-again (the last view).

Jewish voters, Mormon voters and the unaffiliated have positive gaps, +23.8 (95% CI 17.9 to 29.8), +17.5 (95% CI 11.1 to 23.9) and +18.1 (95% CI 14.1 to 22.1) points, with intervals wider than those of the larger groups because one of their two attendance cells is small. The Jewish gap exceeds the white born-again gap by 13.7 (95% CI 7.6 to 19.8) points. A positive gap means that seldom attenders voted more Democratic, and the sign does not say that attendance moves the vote.

The pooled gap is a sum of different gaps and different mixes

The pooled gap can be written as two parts. The within-group part is each group's own seldom-minus-frequent gap, weighted by the group's average share of the two attendance groups. The mix part is the difference between a group's share of seldom attenders and its share of frequent attenders, valued at the group's average Democratic share relative to all voters. The two parts add to the pooled gap exactly, and the split is an accounting of the gap and not a causal one. In 2024 the pooled gap of 27.7 (95% CI 26.0 to 29.4) points, the gap of Chapter 9.1, The god gap in America restricted to voters with a group and an attendance answer, divides into 14.6 (95% CI 11.8 to 17.4) points within groups and 13.1 (95% CI 10.6 to 15.5) points from the mix (Figure 5). Pooled over the five elections the gap is 24.4 (95% CI 23.7 to 25.1) points, of which 11.4 (95% CI 10.3 to 12.4) is within groups and 13.1 (95% CI 12.2 to 13.9) is the mix. Each part is at least 8.0 points in every election, and the mix part is the larger in 3 elections and the within-group part in the others.

Figure 5

Of the 27.7 point gap between seldom and frequent attenders in the 2024 vote, 14.6 points is a gap inside tradition-by-race groups and 13.1 points reflects which groups the two attendance levels contain

Seldom minus frequent attenders in the Democratic share of the two-party presidential vote, percentage points, split into a within-group part and a mix part, 2008 to 2024; 95% intervals

Seldom minus frequent attenders (percentage points)

Show data

Data for Of the 27.7 point gap between seldom and frequent attenders in the 2024 vote, 14.6 points is a gap inside tradition-by-race groups and 13.1 points reflects which groups the two attendance levels contain
Part	Election	Points	Lower	Upper	Voters
Pooled gap	2008	25.6	23.9	27.2	21,124
Gap inside groups	2008	10.7	8.4	12.9	21,124
Mix of groups	2008	14.9	13.1	16.7	21,124
Pooled gap	2012	27.3	25.7	28.8	34,338
Gap inside groups	2012	15.0	12.7	17.2	34,338
Mix of groups	2012	12.3	10.5	14.2	34,338
Pooled gap	2016	18.9	17.4	20.5	37,341
Gap inside groups	2016	8.0	6.1	9.9	37,341
Mix of groups	2016	11.0	9.4	12.5	37,341
Pooled gap	2020	23.5	22.1	25.0	40,502
Gap inside groups	2020	8.7	6.7	10.7	40,502
Mix of groups	2020	14.8	13.3	16.4	40,502
Pooled gap	2024	27.7	26.0	29.4	39,039
Gap inside groups	2024	14.6	11.8	17.4	39,039
Mix of groups	2024	13.1	10.6	15.5	39,039

Each part is at least 8.0 points in every election, and the mix part is the larger in 3 of the five elections. The unaffiliated carry 49.4% of the 2024 within-group part, from a small weekly cell. The lines join the five presidential elections, and the bands are 95% intervals from a bootstrap of respondents.Note: The pooled gap here covers voters with a tradition-by-race group and an attendance answer in the frequent or seldom group, so it can differ by a few tenths of a point from the all-voter gap of the god-gap chapter. The within part weights each group's own gap by its average share of the two attendance groups, and the mix part prices the difference in the groups' shares at each group's average Democratic share; the two parts add to the pooled gap exactly.Note: Standard errors come from 1,000 seeded bootstrap resamples of respondents; the interval is a normal 95% interval around the estimate. The decomposition is an accounting of the gap and not a causal split.Note: The unaffiliated contribute 7.2 of the 14.6 points of the 2024 within-group part, and without them that part is 7.4 points. Their weekly cell has an effective sample of 77 respondents in 2024 against 134, 212 and 208 in 2012, 2016 and 2020, and their gap of +30.1 points in 2024 compares with +20.2, +15.7 and +14.3 in those elections.Source: Cooperative Election Study, presidential years 2008, 2012, 2016, 2020 and 2024; weighted (weight_cumulative)

Figure 6 assigns the parts to groups. White born-again Protestants are 39.0% (95% CI 37.5% to 40.6%) of the voters who attend weekly or more and 9.1% (95% CI 8.6% to 9.7%) of those who attend seldom. They are less Democratic than other voters, so their concentration among weekly attenders adds 10.2 (95% CI 9.5 to 10.9) points to the mix part. Black Protestants are 8.4% (95% CI 7.5% to 9.3%) of weekly attenders and 3.0% (95% CI 2.7% to 3.3%) of seldom attenders and are more Democratic than the average voter, so their contribution to the mix part is −1.8 (95% CI −2.2 to −1.5) points, below zero. White Catholics add 1.5 (95% CI 1.0 to 2.0) points through their own gap and white born-again Protestants 2.3 (95% CI 1.6 to 3.0). Pooled over the five elections, white born-again Protestants contribute 9.2 (95% CI 9.0 to 9.5) points through the mix and the unaffiliated 3.9 (95% CI 3.1 to 4.8) points through their own gap, the largest part of each kind.

Figure 6

White born-again Protestants contribute 10.2 points to the mix part of the 2024 gap and Black Protestants −1.8 points, while the unaffiliated contribute 7.2 points to the within-group part

Each group's contribution to the gap between seldom and frequent attenders in the Democratic share of the two-party vote, percentage points, split into the group's own gap and the mix of attendance groups, 2024 (button: five elections pooled); 95% intervals

Gap inside the group
Mix of the attendance groups

Show data

Data for White born-again Protestants contribute 10.2 points to the mix part of the 2024 gap and Black Protestants −1.8 points, while the unaffiliated contribute 7.2 points to the within-group part
View	Group	Part	Contribution (points)	Lower	Upper	Share of frequent attenders (%)	Share of seldom attenders (%)	Voters in the group
pooled	White born-again Protestants	Gap inside the group	2.3	2.0	2.7	37.3	9.2	27,185
pooled	White born-again Protestants	Mix of the attendance groups	9.2	9.0	9.5	37.3	9.2	27,185
pooled	Unaffiliated	Gap inside the group	3.9	3.1	4.8	2.9	40.5	52,288
pooled	Unaffiliated	Mix of the attendance groups	4.5	3.8	5.2	2.9	40.5	52,288
pooled	White Catholics	Gap inside the group	2.1	1.8	2.3	16.2	13.8	26,811
pooled	White Catholics	Mix of the attendance groups	0.3	0.2	0.4	16.2	13.8	26,811
pooled	Another religion or something else	Gap inside the group	1.5	1.2	1.7	9.7	7.8	12,134
pooled	Another religion or something else	Mix of the attendance groups	0.0	0.0	0.0	9.7	7.8	12,134
pooled	Mormon	Gap inside the group	0.4	0.2	0.5	3.8	0.5	2,375
pooled	Mormon	Mix of the attendance groups	0.8	0.7	0.9	3.8	0.5	2,375
pooled	Jewish	Gap inside the group	0.5	0.4	0.6	1.0	3.0	4,759
pooled	Jewish	Mix of the attendance groups	0.2	0.1	0.2	1.0	3.0	4,759
pooled	Asian or other-race Protestants	Gap inside the group	0.3	0.2	0.4	2.9	1.2	2,812
pooled	Asian or other-race Protestants	Mix of the attendance groups	0.3	0.2	0.4	2.9	1.2	2,812
pooled	Hispanic Protestants	Gap inside the group	0.3	0.1	0.4	3.0	0.8	2,540
pooled	Hispanic Protestants	Mix of the attendance groups	0.2	0.1	0.2	3.0	0.8	2,540
pooled	Hispanic Catholics	Gap inside the group	0.4	0.3	0.6	3.7	3.6	5,659
pooled	Hispanic Catholics	Mix of the attendance groups	0.0	0.0	0.0	3.7	3.6	5,659
pooled	Buddhist or Hindu	Gap inside the group	0.0	−0.1	0.0	0.4	1.2	1,659
pooled	Buddhist or Hindu	Mix of the attendance groups	0.2	0.2	0.2	0.4	1.2	1,659
pooled	Asian or other-race Catholics	Gap inside the group	0.1	0.1	0.2	1.3	0.7	1,523
pooled	Asian or other-race Catholics	Mix of the attendance groups	0.0	0.0	0.0	1.3	0.7	1,523
pooled	Orthodox Christian	Gap inside the group	0.1	0.0	0.1	0.5	0.4	756
pooled	Orthodox Christian	Mix of the attendance groups	0.0	0.0	0.0	0.5	0.4	756
pooled	Muslim	Gap inside the group	0.0	−0.1	0.0	0.5	0.3	526
pooled	Muslim	Mix of the attendance groups	−0.1	−0.1	0.0	0.5	0.3	526
pooled	Black Catholics	Gap inside the group	0.0	0.0	0.1	0.8	0.5	951
pooled	Black Catholics	Mix of the attendance groups	−0.1	−0.2	−0.1	0.8	0.5	951
pooled	White Protestants who are not born-again	Gap inside the group	−0.6	−0.8	−0.4	7.6	13.5	22,658
pooled	White Protestants who are not born-again	Mix of the attendance groups	−0.3	−0.3	−0.2	7.6	13.5	22,658
pooled	Black Protestants	Gap inside the group	0.1	0.0	0.2	8.5	3.0	7,708
pooled	Black Protestants	Mix of the attendance groups	−2.2	−2.3	−2.0	8.5	3.0	7,708
y2024	White born-again Protestants	Gap inside the group	2.3	1.6	3.0	39.0	9.1	5,380
y2024	White born-again Protestants	Mix of the attendance groups	10.2	9.5	10.9	39.0	9.1	5,380
y2024	Unaffiliated	Gap inside the group	7.2	4.7	9.7	3.0	44.8	14,033
y2024	Unaffiliated	Mix of the attendance groups	3.1	1.0	5.3	3.0	44.8	14,033
y2024	White Catholics	Gap inside the group	1.5	1.0	2.0	13.3	12.0	5,448
y2024	White Catholics	Mix of the attendance groups	0.2	0.0	0.3	13.3	12.0	5,448
y2024	Another religion or something else	Gap inside the group	1.4	0.9	1.9	8.5	7.3	2,826
y2024	Another religion or something else	Mix of the attendance groups	0.1	0.0	0.2	8.5	7.3	2,826
y2024	Hispanic Protestants	Gap inside the group	0.3	0.0	0.6	4.3	1.1	570
y2024	Hispanic Protestants	Mix of the attendance groups	0.5	0.3	0.7	4.3	1.1	570
y2024	Jewish	Gap inside the group	0.7	0.3	1.0	1.3	3.0	1,090
y2024	Jewish	Mix of the attendance groups	0.1	0.0	0.2	1.3	3.0	1,090
y2024	Asian or other-race Protestants	Gap inside the group	0.3	0.1	0.6	3.6	1.5	633
y2024	Asian or other-race Protestants	Mix of the attendance groups	0.4	0.2	0.6	3.6	1.5	633
y2024	Mormon	Gap inside the group	0.2	−0.1	0.5	2.9	0.5	416
y2024	Mormon	Mix of the attendance groups	0.4	0.2	0.6	2.9	0.5	416
y2024	Hispanic Catholics	Gap inside the group	0.6	0.2	1.0	4.3	3.5	1,193
y2024	Hispanic Catholics	Mix of the attendance groups	0.0	−0.1	0.0	4.3	3.5	1,193
y2024	Buddhist or Hindu	Gap inside the group	0.1	−0.1	0.3	0.3	1.4	406
y2024	Buddhist or Hindu	Mix of the attendance groups	0.2	0.1	0.3	0.3	1.4	406
y2024	Orthodox Christian	Gap inside the group	0.1	0.0	0.3	0.7	0.3	163
y2024	Orthodox Christian	Mix of the attendance groups	0.1	0.0	0.2	0.7	0.3	163
y2024	Asian or other-race Catholics	Gap inside the group	0.1	−0.1	0.3	1.7	0.9	351
y2024	Asian or other-race Catholics	Mix of the attendance groups	0.0	0.0	0.1	1.7	0.9	351
y2024	Muslim	Gap inside the group	0.1	0.0	0.3	0.6	0.3	126
y2024	Muslim	Mix of the attendance groups	0.0	−0.1	0.0	0.6	0.3	126
y2024	Black Catholics	Gap inside the group	0.0	−0.2	0.1	1.0	0.5	228
y2024	Black Catholics	Mix of the attendance groups	−0.2	−0.3	0.0	1.0	0.5	228
y2024	White Protestants who are not born-again	Gap inside the group	−0.4	−0.8	0.1	6.9	10.7	4,492
y2024	White Protestants who are not born-again	Mix of the attendance groups	−0.2	−0.3	−0.1	6.9	10.7	4,492
y2024	Black Protestants	Gap inside the group	0.0	−0.3	0.4	8.4	3.0	1,684
y2024	Black Protestants	Mix of the attendance groups	−1.8	−2.2	−1.5	8.4	3.0	1,684

White born-again Protestants, who are a large part of weekly attenders and a small part of seldom attenders, raise the pooled gap through the mix, and Black Protestants, who are more Democratic than average and more often weekly attenders, lower it. The unaffiliated carry the largest within-group part. The contributions of all groups add to the pooled gap.Note: The unaffiliated within-group contribution rests on a small number of weekly attenders: the frequent cell has an effective sample of 77 in 2024. Intervals are 95% intervals from the Kish effective sample size and treat the weighted sample as a simple random sample.Note: A group enters when it has at least 30 voters with a frequent or seldom attendance answer in the sample.Source: Cooperative Election Study, presidential years 2008, 2012, 2016, 2020 and 2024; weighted (weight_cumulative)

The unaffiliated need a caution. They are 44.8% (95% CI 43.9% to 45.7%) of seldom attenders and 3.0% (95% CI 2.3% to 3.6%) of weekly attenders, and they contribute 3.1 (95% CI 1.0 to 5.3) points through the mix and 7.2 (95% CI 4.7 to 9.7) points through their own gap. The second is 49.4% (95% CI 38.9% to 59.8%) of the 2024 within-group part, and without the unaffiliated that part is 7.4 (95% CI 6.0 to 8.8) points. It rests on a small weekly cell, with an effective sample of 77 respondents in 2024 against 134, 212 and 208 in 2012, 2016 and 2020. The unaffiliated gap was +0.4 (95% CI −11.5 to 12.3) points in 2008 on an effective weekly sample of 51, then +20.2 (95% CI 11.7 to 28.8), +15.7 (95% CI 8.9 to 22.6) and +14.3 (95% CI 7.5 to 21.0) points, and +30.1 (95% CI 19.0 to 41.3) in 2024. The 2024 estimate is the largest of the five, and its interval overlaps those of 2012, 2016 and 2020.

The pooled gap therefore contains two kinds of difference. One is the difference in vote between seldom and frequent attenders inside a tradition and race, which is small among Black Protestants and below zero in the CES among white Protestants who are not born-again, a sign that other surveys do not reproduce. The other is a difference in who the two attendance groups are. Because Black Protestants are a larger share of frequent attenders than of seldom attenders, they hold the pooled gap below what the other groups alone would give. The born-again rule also builds part of the white born-again contribution into the coding, because the group is defined by an answer that marks religious commitment.

In the GSS white RELTRAD evangelical Protestants leaned Democratic in the 1970s

The CES begins in 2008. The GSS asks Protestants which denomination they belong to, which allows RELTRAD from 1972 to 2022 (68,896 respondents placed; the 2024 GSS has no detailed denomination). This chapter codes it from the denomination and open-ended answers with the published syntax of the scheme, as the methods note describes. Figure 7 follows the Democratic share of party identifiers in five windows of survey years. Among white RELTRAD evangelical Protestants it fell from 61.0% (95% CI 59.0% to 63.0%) in 1972 to 1984 to 43.8% in 1985 to 1996, 35.3% in 1998 to 2008 and 30.6% (95% CI 28.1% to 33.1%) in 2010 to 2018. Black Protestants stayed between 88.5% and 91.1% in every window, and white RELTRAD mainline Protestants between 40.9% and 45.4%. Catholics, of any race, fell from 72.4% (95% CI 70.7% to 74.1%) to 56.9% (95% CI 55.1% to 58.8%) and were at 58.3% (95% CI 55.7% to 60.9%) in 2010 to 2018. The GSS race item counts respondents of Hispanic origin who say white as white: 7.7% of white RELTRAD evangelical respondents in 2010 to 2018 are of Hispanic origin, and the non-Hispanic Democratic share is 28.8% (95% CI 26.3% to 31.4%) against 30.6%.

Figure 7

In the GSS the Democratic share of party identifiers among white RELTRAD evangelical Protestants fell from 61.0% in 1972 to 1984 to 30.6% in 2010 to 2018, while Black Protestants stayed near 91.1%

Democratic share of Democratic and Republican identifiers (leaners included) by RELTRAD tradition, for all members and for frequent and seldom attenders, five survey windows from 1972 to 2022, percent; 95% intervals (button: the ANES presidential vote, 1980 to 2020)

Democratic share of party identifiers (%)

All members
Frequent attenders
Seldom attenders

Show data

Data for In the GSS the Democratic share of party identifiers among white RELTRAD evangelical Protestants fell from 61.0% in 1972 to 1984 to 30.6% in 2010 to 2018, while Black Protestants stayed near 91.1%
Survey	Group	Group of attenders	Survey years	Democratic share (%)	Lower	Upper	Respondents
GSS	Black Protestants (RELTRAD)	All members	1972 to 1984	88.5	86.5	90.5	1,537
GSS	Black Protestants (RELTRAD)	All members	1985 to 1996	88.5	86.5	90.4	1,576
GSS	Black Protestants (RELTRAD)	All members	1998 to 2008	90.9	89.0	92.7	1,266
GSS	Black Protestants (RELTRAD)	All members	2010 to 2018	91.1	88.6	93.6	752
GSS	Black Protestants (RELTRAD)	All members	2021 to 2022	88.5	83.5	93.5	320
GSS	Black Protestants (RELTRAD)	Frequent attenders	1972 to 1984	89.1	85.9	92.3	586
GSS	Black Protestants (RELTRAD)	Frequent attenders	1985 to 1996	90.2	87.2	93.1	592
GSS	Black Protestants (RELTRAD)	Frequent attenders	1998 to 2008	91.0	88.1	93.9	532
GSS	Black Protestants (RELTRAD)	Frequent attenders	2010 to 2018	93.1	89.5	96.7	299
GSS	Black Protestants (RELTRAD)	Frequent attenders	2021 to 2022	91.0	82.9	99.2	126
GSS	Black Protestants (RELTRAD)	Seldom attenders	1972 to 1984	89.4	86.0	92.8	512
GSS	Black Protestants (RELTRAD)	Seldom attenders	1985 to 1996	86.6	83.0	90.2	496
GSS	Black Protestants (RELTRAD)	Seldom attenders	1998 to 2008	89.8	86.2	93.4	383
GSS	Black Protestants (RELTRAD)	Seldom attenders	2010 to 2018	90.0	85.3	94.7	236
GSS	Black Protestants (RELTRAD)	Seldom attenders	2021 to 2022	83.5	75.2	91.8	138
GSS	Catholics	All members	1972 to 1984	72.4	70.7	74.1	3,635
GSS	Catholics	All members	1985 to 1996	56.9	55.1	58.8	3,813
GSS	Catholics	All members	1998 to 2008	56.9	55.0	58.9	3,354
GSS	Catholics	All members	2010 to 2018	58.3	55.7	60.9	2,117
GSS	Catholics	All members	2021 to 2022	53.2	49.1	57.3	1,188
GSS	Catholics	Frequent attenders	1972 to 1984	71.3	68.9	73.8	1,788
GSS	Catholics	Frequent attenders	1985 to 1996	57.1	54.2	60.0	1,552
GSS	Catholics	Frequent attenders	1998 to 2008	53.6	50.3	56.9	1,210
GSS	Catholics	Frequent attenders	2010 to 2018	54.1	49.3	58.9	642
GSS	Catholics	Frequent attenders	2021 to 2022	52.0	44.0	60.0	315
GSS	Catholics	Seldom attenders	1972 to 1984	72.1	69.3	74.9	1,349
GSS	Catholics	Seldom attenders	1985 to 1996	57.9	55.1	60.7	1,576
GSS	Catholics	Seldom attenders	1998 to 2008	58.7	55.8	61.5	1,555
GSS	Catholics	Seldom attenders	2010 to 2018	60.6	57.0	64.2	1,062
GSS	Catholics	Seldom attenders	2021 to 2022	53.4	48.2	58.5	743
GSS	White evangelical Protestants (RELTRAD)	All members	1972 to 1984	61.0	59.0	63.0	3,099
GSS	White evangelical Protestants (RELTRAD)	All members	1985 to 1996	43.8	42.0	45.5	3,902
GSS	White evangelical Protestants (RELTRAD)	All members	1998 to 2008	35.3	33.4	37.2	3,076
GSS	White evangelical Protestants (RELTRAD)	All members	2010 to 2018	30.6	28.1	33.1	1,855
GSS	White evangelical Protestants (RELTRAD)	All members	2021 to 2022	16.0	12.4	19.6	834
GSS	White evangelical Protestants (RELTRAD)	Frequent attenders	1972 to 1984	55.0	52.0	58.1	1,415
GSS	White evangelical Protestants (RELTRAD)	Frequent attenders	1985 to 1996	37.4	34.9	40.0	1,684
GSS	White evangelical Protestants (RELTRAD)	Frequent attenders	1998 to 2008	26.5	24.0	29.0	1,517
GSS	White evangelical Protestants (RELTRAD)	Frequent attenders	2010 to 2018	23.2	20.0	26.4	941
GSS	White evangelical Protestants (RELTRAD)	Frequent attenders	2021 to 2022	13.7	9.2	18.3	488
GSS	White evangelical Protestants (RELTRAD)	Seldom attenders	1972 to 1984	67.4	64.3	70.5	1,217
GSS	White evangelical Protestants (RELTRAD)	Seldom attenders	1985 to 1996	49.3	46.5	52.0	1,522
GSS	White evangelical Protestants (RELTRAD)	Seldom attenders	1998 to 2008	45.8	42.4	49.3	1,020
GSS	White evangelical Protestants (RELTRAD)	Seldom attenders	2010 to 2018	40.2	35.5	44.9	606
GSS	White evangelical Protestants (RELTRAD)	Seldom attenders	2021 to 2022	18.4	11.6	25.3	231
GSS	White mainline Protestants (RELTRAD)	All members	1972 to 1984	45.4	43.6	47.2	4,018
GSS	White mainline Protestants (RELTRAD)	All members	1985 to 1996	42.2	40.3	44.1	3,117
GSS	White mainline Protestants (RELTRAD)	All members	1998 to 2008	41.6	39.3	43.9	2,215
GSS	White mainline Protestants (RELTRAD)	All members	2010 to 2018	40.9	37.3	44.5	1,091
GSS	White mainline Protestants (RELTRAD)	All members	2021 to 2022	45.0	39.9	50.1	641
GSS	White mainline Protestants (RELTRAD)	Frequent attenders	1972 to 1984	38.6	35.5	41.8	1,224
GSS	White mainline Protestants (RELTRAD)	Frequent attenders	1985 to 1996	40.4	36.8	44.0	877
GSS	White mainline Protestants (RELTRAD)	Frequent attenders	1998 to 2008	34.9	30.9	38.9	674
GSS	White mainline Protestants (RELTRAD)	Frequent attenders	2010 to 2018	44.4	37.8	50.9	344
GSS	White mainline Protestants (RELTRAD)	Frequent attenders	2021 to 2022	36.9	28.0	45.8	174
GSS	White mainline Protestants (RELTRAD)	Seldom attenders	1972 to 1984	51.0	48.5	53.5	2,008
GSS	White mainline Protestants (RELTRAD)	Seldom attenders	1985 to 1996	44.4	41.7	47.2	1,566
GSS	White mainline Protestants (RELTRAD)	Seldom attenders	1998 to 2008	45.5	42.2	48.9	1,100
GSS	White mainline Protestants (RELTRAD)	Seldom attenders	2010 to 2018	40.8	35.8	45.9	541
GSS	White mainline Protestants (RELTRAD)	Seldom attenders	2021 to 2022	47.0	40.5	53.5	393
ANES	Black Protestants	All members	1980 to 1988	91.7	88.6	94.9	303
ANES	Black Protestants	All members	1992 to 2000	93.8	91.0	96.6	328
ANES	Black Protestants	All members	2008 to 2012	97.6	96.2	99.1	918
ANES	Black Protestants	All members	2016 to 2020	93.1	90.2	96.1	512
ANES	White Catholics	All members	1980 to 1988	42.9	39.3	46.5	725
ANES	White Catholics	All members	1992 to 2000	47.0	43.5	50.6	843
ANES	White Catholics	All members	2008 to 2012	46.6	42.2	51.0	742
ANES	White Catholics	All members	2016 to 2020	42.5	39.4	45.7	1,427
ANES	White Protestants who are not born-again	All members	1980 to 1988	30.9	27.8	33.9	865
ANES	White Protestants who are not born-again	All members	1992 to 2000	44.1	40.6	47.5	871
ANES	White Protestants who are not born-again	All members	2008 to 2012	44.4	39.9	49.0	702
ANES	White Protestants who are not born-again	All members	2016 to 2020	47.0	43.6	50.4	1,331
ANES	White born-again Protestants	All members	1980 to 1988	30.1	26.8	33.5	710
ANES	White born-again Protestants	All members	1992 to 2000	29.3	25.9	32.6	787
ANES	White born-again Protestants	All members	2008 to 2012	21.9	18.6	25.2	854
ANES	White born-again Protestants	All members	2016 to 2020	19.1	16.5	21.8	1,437

White RELTRAD evangelical Protestants moved from a Democratic majority to a Republican one, white RELTRAD mainline Protestants and Black Protestants stayed in place, and Catholics fell and then held. Each panel is one tradition, each point a pooled window of survey years placed at its midpoint, and each band a 95% interval. The shaded 2021 to 2022 window includes web interviews.Note: GSS view: tradition follows RELTRAD (Steensland and others 2000) from the GSS denomination items; the 2024 GSS has no detailed denomination, so the series ends in 2022. White evangelical and white mainline Protestants are RELTRAD evangelical and mainline respondents whose GSS race is white, which includes white respondents of Hispanic origin. Catholics are all Catholics, since the GSS records Hispanic origin only from 2000.Note: Frequent attenders attend nearly every week or more; seldom attenders a few times a year or never. The window 2021 to 2022 mixes web and in-person interviews: 87.8% of 2021 respondents were interviewed on the web. Respondents that RELTRAD cannot place are left out, 7.6% of respondents in 2021 to 2022. A cell is drawn only when it has at least 30 respondents and an effective sample of at least 30. Intervals are 95% intervals from the Kish effective sample size and treat the weighted sample as a simple random sample.Note: ANES view: reported two-party presidential vote, with the born-again rule of the CES. The born-again item is asked in every presidential-year study from 1980 to 2020 except 2004, and a respondent with no answer is in neither white Protestant group. The 2012 and 2016 studies add web samples and the 2020 study was collected online, by telephone and by video, so the last windows mix interview modes with the in-person studies of earlier years; there are no mode weights for 2020. The ANES weight is 1 for every respondent from 1980 to 1990. The 2024 religion codes of the cumulative file are unusable, so the series ends in 2020.Source: General Social Survey, 1972-2022, and ANES Cumulative Data File, 1980-2020; weighted (weight_use and VCF0009z, each rescaled to average one within each survey year)

Attendance separated white RELTRAD evangelical Protestants before the group moved. In 1972 to 1984 frequent attenders were 55.0% (95% CI 52.0% to 58.1%) Democratic and seldom attenders 67.4% (95% CI 64.3% to 70.5%), a gap of +12.4 (95% CI 8.0 to 16.8) points, and the gap was +17.1 (95% CI 11.4 to 22.7) in 2010 to 2018. Its interval lies above zero in every window except the last, where it includes zero (+4.7 (95% CI −3.5 to 12.9)). The Black Protestant gap has an interval that includes zero in all five windows. The white RELTRAD mainline gap was +12.3 (95% CI 8.3 to 16.4) points in 1972 to 1984 and −3.5 (95% CI −11.8 to 4.8) in 2010 to 2018, so frequent mainline attenders were less Democratic than seldom attenders early and by the 2010s the gap is within sampling error of zero. The 2021 to 2022 window, at +10.1 (95% CI −0.9 to 21.2), does not continue that path.

The last window includes web interviews and its estimates differ with the mode. White RELTRAD evangelical Protestants were 16.0% (95% CI 12.4% to 19.6%) Democratic across 2021 to 2022. In 2022, in-person respondents gave 9.2% (95% CI 2.2% to 16.3%) and web respondents 12.4% (95% CI 4.3% to 20.5%), and the 2021 survey, 87.8% of it on the web, gave 21.3% (95% CI 16.5% to 26.1%). RELTRAD also leaves 7.6% of respondents in that window unplaced, 5.4% of respondents being Protestants with no denomination who attend less than monthly and 1.6% being respondents who say only Christian or interdenominational, a gap in the syntax that [4] describe. Counting every white respondent that RELTRAD cannot place as evangelical gives 23.5% (95% CI 20.0% to 27.0%) Democratic, and counting only those who say Christian or interdenominational gives 18.0% (95% CI 14.4% to 21.6%). Even the larger estimate lies below the estimate for 2010 to 2018.

The ANES cumulative file measures the presidential vote and not party identification, and its born-again item, asked in every presidential-year study from 1980 to 2020 except 2004, allows the CES rule (the second view of Figure 7). White born-again Protestants voted 30.1% (95% CI 26.8% to 33.5%) Democratic in 1980 to 1988 and 19.1% (95% CI 16.5% to 21.8%) in 2016 to 2020. By party identification the GSS puts white RELTRAD evangelical Protestants at 61.0% in the first window and the ANES puts white born-again Protestants at 44.0% (95% CI 40.8% to 47.1%), so the gap between the GSS party share and the ANES vote share has two sources, the coding of the group and the difference between identification and vote. White Protestants who are not born-again voted 30.9% (95% CI 27.8% to 33.9%) Democratic in 1980 to 1988 and 47.0% (95% CI 43.6% to 50.4%) in 2016 to 2020, a rise whose intervals do not overlap, while their party identification is flat at 42.5% and 42.0%, as the GSS series is. Black Protestants were at 91.7% and 93.1% and white Catholics at 42.9% and 42.5%. The 2012 and 2016 ANES studies add web samples and the 2020 study was collected online, by telephone and by video, so the last windows mix interview modes.

The sign of the gap among white Protestants who are not born-again depends on the survey and the coding

The CES rule puts a white Protestant in the born-again group when the respondent calls themselves born-again or evangelical. In the GSS from 2006 to 2018, the two codings agree for 61.3% of white Protestants with a born-again answer and a party identification, counting respondents RELTRAD cannot place as disagreeing and leaving out pure independents, and for 70.7% of those RELTRAD places in the evangelical or mainline tradition. Among white Protestants RELTRAD places, with no condition on party identification, 39.0% of those who are not born-again are RELTRAD evangelical and 22.3% of those who are born-again are RELTRAD mainline, so the CES groups are not the denominational ones. Figure 8 shows what that does. The first tab compares 2024 levels in the CES and the ANES with the born-again rule applied in both. The ANES puts white Protestants who are not born-again at 56.1% (95% CI 49.9% to 62.3%) Democratic and the CES at 43.0% (95% CI 41.0% to 45.0%), a difference of +13.2 (95% CI 6.7 to 19.7) points, and the major-group summary of the ANES gives 57.6% (95% CI 51.0% to 64.1%) for its mainline category. The ANES national Democratic share, 54.4% (95% CI 52.2% to 56.7%), is above the CES share, 49.0% (95% CI 48.2% to 49.7%), and the certified count, 49.2%, so part of the difference belongs to the survey. The ANES also places each of six groups above the CES and orders them the same way.

Figure 8

The ANES and the CES differ by +13.2 points for white Protestants who are not born-again in 2024, and only the CES puts their attendance gap below zero

Two checks on the CES tradition-by-race coding: 2024 levels in the CES and the ANES Time Series, and the attendance gap for white Protestants under each survey and coding

In 2024 the ANES places each of six groups above the CES, most for white Protestants who are not born-again and Hispanic Catholics, and the two surveys order the groups identically

Democratic share of the two-party presidential vote, reported voters, percent, 2024; white Protestants are split by the born-again rule in both surveys; 95% intervals

CES
ANES

White Protestants who are not born-again are 56.1% Democratic in the ANES and 43.0% in the CES, and the intervals do not overlap. Each dot is a weighted share with a 95% interval, in the 2024 CES and the 2024 ANES Time Series.

Show data

Data for The ANES and the CES differ by +13.2 points for white Protestants who are not born-again in 2024, and only the CES puts their attendance gap below zero, Levels in 2024
Group	Survey	Democratic share (%)	Lower	Upper	Respondents
White born-again Protestants	CES	14.5	13.2	15.7	5,987
White born-again Protestants	ANES	20.9	15.7	26.0	497
White Protestants who are not born-again	CES	43.0	41.0	45.0	4,979
White Protestants who are not born-again	ANES	56.1	49.9	62.3	478
Black Protestants	CES	83.1	80.3	86.0	1,889
Black Protestants	ANES	90.2	84.0	96.3	179
White Catholics	CES	37.1	35.4	38.9	5,974
White Catholics	ANES	44.4	38.2	50.5	540
Hispanic Catholics	CES	57.0	52.5	61.5	1,351
Hispanic Catholics	ANES	71.5	59.9	83.1	123
Unaffiliated	CES	70.9	69.7	72.2	14,311
Unaffiliated	ANES	78.6	74.6	82.5	854

The first tab compares levels in 2024 and the second compares attendance gaps. Under the CES born-again rule and under the denominational codings of the GSS (RELTRAD) and the ANES (its major-group summary), the evangelical or born-again group shows a positive gap, and the sign of the gap in the mainline or not-born-again group depends on the survey.Note: The CES rule puts white Protestants who call themselves born-again or evangelical in the evangelical group and the others in the mainline group. The GSS RELTRAD rows use denomination, the ANES summary rows use the ANES major-group religion summary (V241445x), and the born-again rows in the GSS and the ANES apply the CES rule to those surveys. This chapter has not checked the ANES summary against RELTRAD. In the GSS, 61.3% of white Protestants with a born-again answer and a party identification fall in the same group under RELTRAD and the born-again rule, counting respondents RELTRAD cannot place as disagreeing and leaving out pure independents; among those RELTRAD places in the evangelical or mainline tradition the share is 70.7%.Note: Among white Protestants whom RELTRAD places in the evangelical or mainline tradition, 39.0% of those who are not born-again are RELTRAD evangelical and 22.3% of those who are born-again are RELTRAD mainline.Note: GSS rows cover the surveys of 2006 to 2018, in all interview modes those surveys used, which include some telephone interviews; the CES rows cover the five presidential elections, and the ANES rows 2024, which mixed in-person, web and mail-return samples under one post-election weight. Frequent in the GSS is nearly weekly or more and in the ANES weekly or almost weekly, one answer wider than the CES. A cell is drawn only when it has at least 30 respondents and an effective sample of at least 30. Intervals are 95% intervals from the Kish effective sample size and treat the weighted sample as a simple random sample.Source: Cooperative Election Study, 2008-2024; General Social Survey, 2006-2018; ANES 2024 Time Series; weighted (weight_cumulative, weight_use and V240107b)

The second tab gives the attendance gap for the evangelical and the mainline group of white Protestants under six combinations of survey and coding. The gap in the evangelical group has a positive point estimate in every row: +10.1 (95% CI 8.7 to 11.4) points in the CES vote, +19.3 (95% CI 14.6 to 24.1) in the GSS under RELTRAD and +11.9 (95% CI 0.1 to 23.8) in the ANES under its own summary, and its interval lies above zero in every row except the ANES born-again row. The sizes differ across surveys. The gap in the mainline group is below zero in the CES, −5.4 (95% CI −7.5 to −3.4) points for the vote and −3.7 (95% CI −5.8 to −1.7) for party identification, and its interval includes zero under the denominational codings of the GSS, +2.8 (95% CI −3.8 to 9.4), and the ANES, +11.6 (95% CI −3.4 to 26.6). Applying the CES born-again rule to the GSS gives +5.7 (95% CI −0.5 to 12.0) and to the ANES +6.2 (95% CI −7.9 to 20.3), so the rule alone does not reproduce the CES sign. The data do not say whether the CES result comes from its sample, from the mode of interview or from the rule, and the gap among white Protestants who are not born-again is the least settled estimate in this chapter.

What these data cannot show

Each survey interviews a different sample, so the gaps compare groups and say nothing about any voter changing. Religious behavior and identity respond to politics as well as the reverse: panel studies find that partisan identity shapes whether people identify with a religion and how often they attend [5] [6]. A gap between seldom and frequent attenders, or a difference between traditions, can therefore reflect people moving between groups, and the decomposition splits an observed gap into parts without saying what would happen if the mix of attenders changed.

Tradition in the CES is a proxy. The born-again rule ties the white born-again group to an answer that marks religious commitment, and the group of white Protestants who are not born-again is whoever is left. The CES cannot separate Baptist, Methodist, Lutheran and non-denominational respondents, cannot place Black Protestants by denomination, and cannot tell Pentecostal from other Hispanic Protestants. The GSS codes tradition by denomination but stops in 2022, records Hispanic origin only from 2000, and leaves 5.3% of respondents in 2010 to 2018 unplaced. The ANES 2024 rows of the coding check have wider intervals than the CES rows, and the ANES sample is too small to give a Hispanic Protestant estimate at all.

The levels depend on the survey and the weight. The CES estimates use the cumulative-file weight because it is the one CES weight defined for all five elections; the validated-vote weight exists only for 2018 to 2024. In 2024 it gives 34.1% (95% CI 32.4% to 35.8%) for white Catholics against 37.1%, 13.2% (95% CI 11.9% to 14.4%) for white born-again Protestants against 14.5% and 59.6% (95% CI 53.9% to 65.2%) for Hispanic Catholics against 57.0%. The CES national Democratic share is 49.0% with the cumulative-file weight and 49.1% with the validated-vote weight, against a certified 49.2%, and agreement at the national level does not show that a survey is accurate inside a tradition and race. The ANES group levels exceed the CES levels by the most for white Protestants who are not born-again and Hispanic Catholics, and by more than the two national shares differ. Small groups carry intervals as wide as the Muslim estimate, and the figures draw those intervals; the only cells removed are those under the floor.

The sample is voters who reported a Democratic or Republican vote, so nonvoters, third-party voters and respondents who did not answer are outside it. The chapter has no measure of belief, congregation, region or age, and nothing in it identifies why Black Protestants and white born-again Protestants vote as they do. The growth of the unaffiliated and their subtypes belong to Chapter 9.2, The rise of the nones (planned), and the link between attendance and moral attitudes to Chapter 9.5, Two moralities (planned).

Data sources

Cooperative Election StudyCumulative Common Content, presidential years 2008, 2012, 2016, 2020 and 2024; religion, race, born-again, attendance, vote and partyYears: 2006-2024 (every year; self-reported turnout is available in the 10 even years) · Weights: weight_cumulative; vvweight for the 2024 sensitivity check · License: See the provider's terms of use.
General Social SurveyCumulative data 1972 to 2022 from the raw file: relig, denom (detailed), other, race, hispanic, attend, partyid, reborn; mode joined from the raw fileYears: 1972-2024 (35 study years) · Weights: weight_use, rescaled within year · License: See the provider's terms of use.
ANES Cumulative Data FileCumulative Data File, presidential-year studies 1980 to 2020: religion, born-again, race, party identification and presidential vote; the 2024 religion codes are unusableYears: 1948-2024 (33 study years; even years only after 2004; no 2006, 2010 or 2014 study) · Weights: weight (VCF0009z) · License: Free download after accepting the ANES terms of use; cite the release.
ANES 2024 Time Series2024 Time Series: major-group religion summary, race, attendance, born-again, 2024 vote and turnoutYears: 2024 (pre-election and post-election interviews) · Weights: V240107b · License: Free download after accepting the ANES terms of use; cite the release.
State presidential returnsState presidential returns 1976 to 2024, used only for the certified national two-party share in 2024Years: 1976-2024 (13 elections) · Weights: none (full counts) · License: See the provider's terms of use.

All data sources

Methods

Sample. Cooperative Election Study (CES) cumulative file, presidential years 2008, 2012, 2016, 2020 and 2024: 188,145 respondents who reported voting for the Democratic or the Republican presidential candidate (third-party and other votes are excluded), and 227,049 respondents who identify with or lean toward a party. Outcomes are the Democratic share of the reported two-party presidential vote and the Democratic share of Democratic and Republican identifiers (CES pid7 1 to 3 against 5 to 7, leaners with their party). Tradition and race in the CES. The cumulative file has religion (1 Protestant, 2 Catholic, 3 Mormon, 4 Orthodox, 5 Jewish, 6 Muslim, 7 Buddhist, 8 Hindu, 9 atheist, 10 agnostic, 11 nothing in particular, 12 something else), race (race5: white non-Hispanic, Black, Hispanic, Asian, other) and the born-again or evangelical self-description (relig_bornagain). White Protestants who say yes are white born-again Protestants and those who say no are white Protestants who are not born-again. Black, Hispanic and Asian or other-race Protestants and Black, Hispanic, white and Asian or other-race Catholics are separate groups. Codes 9 to 11 are the unaffiliated. There is no denomination item, so the born-again rule is a proxy for the evangelical and mainline traditions, and the chapter calls the CES groups by the rule and keeps the names evangelical and mainline for the denominational codings of the GSS and the ANES. Attendance is the six-point item church_attend_6pt: frequent is once a week or more than once a week (codes 5 and 6) and seldom is never, seldom or a few times a year (codes 1 to 3); monthly attenders (code 4) are in neither group. Weights and pooling. CES estimates use weight_cumulative, which sums to the same total in every year over all respondents, so pooled estimates over the five elections count each election by its weighted voters. It is used instead of the validated-vote weight vvweight because it is the only CES weight defined for all five elections (vvweight exists for 2018 to 2024, and it also requires a validated turnout record); the text gives the 2024 shares of three groups and of all voters under vvweight. Pooled GSS windows use weight_use rescaled to average one within each survey year. ANES 2024 estimates use the post-election weight V240107b. Estimators and intervals. A share is a weighted mean of a 0/1 variable, and a gap or a change is the difference of two independent shares with the standard errors combined in quadrature; a difference between two changes combines the four standard errors, and a difference between two gaps combines the two. Every interval is a 95% interval from the Kish effective sample size with a normal approximation. The CES is an opt-in online panel, so its intervals are model based and cover sampling error and not the recruitment of respondents, and none of the intervals follows the clustering of a sampling design. Suppression: a cell is published only with at least 30 respondents and an effective sample of at least 30. The figures draw the interval of every published cell, and the tables give its width. Decomposition. With p_gS and p_gF the shares of group g among seldom and frequent attenders and m_gS and m_gF its Democratic percentages, the seldom-minus-frequent gap equals the sum over groups of pbar_g (m_gS minus m_gF) plus the sum over groups of (p_gS minus p_gF) mbar_g, where pbar_g and mbar_g are averages of the two cells. The first sum is the within-group part and the second the mix part; subtracting the overall mean from mbar_g changes no total and gives each group's contribution. Standard errors are the standard deviation over 1,000 resamples of respondents with replacement (seed 12345), and the interval is a normal 95% interval around the estimate. The within-group part without the unaffiliated and the unaffiliated share of the within-group part use the same resamples. GSS RELTRAD. Tradition follows Steensland and others (2000), coded from the posted syntax of Jerry Park for the Association of Religion Data Archives (6 August 2018), which implements the scheme: relig (religious preference), the detailed denomination denom (the file's denom_7222, asked 1972 to 2022; the 2024 GSS has only a seven-group denom and no other-denomination code, so 2024 cannot be coded), the open-ended other-denomination code other, race and attend. The variable fund is not used by the syntax. Black Protestant: the National Baptist Conventions, the African Methodist Episcopal churches, a list of Black sectarian other codes, and Black respondents in other Baptist, American Baptist, Southern Baptist, Baptist-do-not-know-which, other Methodist and Methodist-do-not-know-which groups. Evangelical: a list of other codes, the American Baptist Association, Southern Baptist and other Baptist groups, white Baptist-do-not-know-which, white other Methodist, Missouri, Wisconsin and other Lutheran synods, the Presbyterian Church in America, and Protestants with no denomination who attend at least monthly. Mainline: a list of other codes, United Methodist, white Methodist-do-not-know-which, American Baptist Churches USA (white), the other Lutheran and Presbyterian groups and Episcopalians. Catholic, Jewish, other faith and unaffiliated follow relig and a list of other codes; Jacobite Apostolics are coded Orthodox. Respondents the syntax cannot place are left out of every tradition: Protestants with no denomination who attend less than monthly, respondents who answer only Christian or interdenominational, and other codes 161 and 185. The syntax's own comment says codes 161 and 185 are dropped but its order of statements leaves them classified; this chapter drops them. 68,896 respondents are placed. Stetzer and Burge (2016) describe the omission of generic Christian and interdenominational answers from the evangelical tradition. This chapter keeps the syntax as written and bounds the effect for 2021 to 2022: the text gives the white evangelical Democratic share when every white respondent RELTRAD cannot place (Protestants with no denomination who attend less than monthly, and those who say only Christian or interdenominational) is counted as evangelical, and when only those who say Christian or interdenominational are. Tradition by race in the GSS: white evangelical and white mainline Protestants are RELTRAD evangelical and mainline respondents whose race is white (race 1), which includes white respondents of Hispanic origin (Hispanic origin is recorded from 2000, and the text gives the non-Hispanic share for 2010 to 2018); Black Protestants are the RELTRAD Black Protestant category; Catholics are all Catholics; Black respondents that RELTRAD places in the evangelical or mainline traditions and respondents of other races in those two traditions belong to no group. Windows pool survey years (1972 to 1984, 1985 to 1996, 1998 to 2008, 2010 to 2018, 2021 to 2022). Party is partyid 0 to 2 against 4 to 6, and attendance groups are frequent (attend 6 to 8, nearly every week or more) and seldom (attend 0 to 3). The born-again rule in the GSS uses reborn (1 yes, 2 no) among white respondents who say Protestant, 2006 to 2018. The agreement between RELTRAD and the born-again rule is computed on white Protestants with a born-again answer and a party identification (pure independents out; respondents RELTRAD cannot place count as disagreeing) and, separately, among those RELTRAD places in the evangelical or mainline tradition; the cross-check of each group by the other rule uses white Protestants with a born-again answer whom RELTRAD places, with no condition on party identification. ANES 2024 Time Series. Religion is the ANES major-group summary V241445x (mainline, evangelical, Black Protestant, undifferentiated Protestant, Roman Catholic, other Christian, Jewish, other religion, not religious); this chapter has not checked its rules against RELTRAD. Race is V241501x. In the coding check the ANES evangelical and mainline groups are white respondents in those two ANES categories. For the comparison of levels with the CES, white Protestants are split by the born-again rule (V241442) among white respondents in the four ANES Protestant categories, as in the CES; Black and Hispanic Protestants are Black and Hispanic respondents in any Protestant category, and Catholics are split by race. Attendance is V241440 (frequent: every week or almost every week; seldom: a few times a year or never, with V241439 = no counted as never). A reported voter has V242095x = voted and V242096x = Harris or Trump; 3,928 respondents meet this. The 2024 study mixes face-to-face, web and mail-return samples under one weight and is not split by mode. ANES Cumulative Data File. Respondents in the presidential-year studies of 1980 to 2020 (the 2024 religion codes of the cumulative file are unusable); the midterm studies of the file carry no presidential vote. Groups follow the CES rule: white non-Hispanic Protestants (religion, race_4cat) who say they are born-again (born_again, asked in every presidential-year study from 1980 to 2020 except 2004) and who do not, Black non-Hispanic Protestants, and white non-Hispanic Catholics. A respondent without a born-again answer is in neither white Protestant group. Windows are 1980 to 1988, 1992 to 2000, 2008 to 2012 and 2016 to 2020. The vote is the reported two-party presidential vote of voters, and party identification is pid_7pt 1 to 3 against 5 to 7 among all respondents. The weight VCF0009z is 1 for every respondent from 1980 to 1990 and is rescaled to average one within each study among voters or among party identifiers. The 2012 and 2016 studies add web samples and the 2020 study was collected online, by telephone and by video, with no mode weights for 2020, so the last windows mix interview modes. Survey mode. The CES is online in every year. GSS survey years from 2021 include web interviews; the 2021 to 2022 window is flagged in the figure and the text, and the 2022 split by mode is reported for white RELTRAD evangelical Protestants. The GSS rows of the coding check use the surveys of 2006 to 2018 in all the interview modes those surveys used, which include some telephone interviews.

References

Steensland, Brian; Park, Jerry Z.; Regnerus, Mark D.; Robinson, Lynn D.; Wilcox, W. Bradford; Woodberry, Robert D. (2000). The Measure of American Religion: Toward Improving the State of the Art. Social Forces 79(1): 291-318. doi:10.2307/2675572
Kish, Leslie (1965). Survey Sampling. John Wiley and Sons.
Lincoln, C. Eric; Mamiya, Lawrence H. (1990). The Black Church in the African American Experience. Duke University Press.
Stetzer, Ed; Burge, Ryan P. (2016). Reltrad Coding Problems and a New Repository. Politics and Religion 9(1): 187-190. doi:10.1017/S1755048315000929
Margolis, Michele F. (2018). From Politics to the Pews: How Partisanship and the Political Environment Shape Religious Identity. University of Chicago Press.
Campbell, David E.; Layman, Geoffrey C.; Green, John C.; Sumaktoyo, Nathanael G. (2018). Putting Politics First: The Impact of Politics on American Religious and Secular Orientations. American Journal of Political Science 62(3): 551-565. doi:10.1111/ajps.12365

How to cite

@misc{singh2026religioustraditionandrace,
  author = {Singh, Vishal},
  title = {Religious tradition and race},
  year = {2026},
  url = {https://vishalsingh.org/politics/religious-tradition-and-race},
  note = {In The Public Mind. Published 3 October 2026.}
}

Two groups that attend often sit at opposite ends of the vote

The five largest Christian groups moved toward the Republican candidate by different amounts

Attendance relates to the vote in different directions inside each group

The pooled gap is a sum of different gaps and different mixes

In the GSS white RELTRAD evangelical Protestants leaned Democratic in the 1970s

The sign of the gap among white Protestants who are not born-again depends on the survey and the coding

What these data cannot show

Data sources

Methods

References

How to cite

Reading paths