By contrast, both the 2007 and 2014 surveys were conducted entirely by phone. In both of those surveys, we recruited respondents using random-digit dialing, and the survey was administered by live interviewers who asked the questions and recorded the answers.[100. numoffset=”100″ It is no longer feasible to conduct high-quality telephone surveys at the scale required by the Religious Landscape Study (RLS) because response rates to telephone polls have declined too far and costs have risen too high. Today, web and paper surveys conducted among respondents recruited using address-based sampling (ABS) techniques produce far higher response rates and provide higher-quality data than do telephone surveys. That is why we made the switch from an interviewer-administered mode (telephone) to a mainly self-administered mode (online and on paper), despite the complications the \u201cmode switch\u201d introduces for comparing results from the new survey with the previous surveys.]<\/p>\n\n\n\n
Respondents answer some survey questions differently when they are talking with an interviewer (as in the 2007 and 2014 surveys) than when they participate in surveys online or on paper (as most respondents in the new survey did). This means that for some<\/em> survey questions, the results we obtained in previous surveys (using an interviewer-administered mode) cannot be directly compared with the results from the new survey that we conducted using mainly self-administered modes, even if the wording of the questions is identical.[101. In addition to the different responses they might elicit from some individuals, surveys conducted using different modes might yield a different mix of respondents to begin with. That is, the mix of people who might agree to participate in a telephone survey might be different in important ways than the sample of people who agree to participate in a web\/paper survey. Over time, the number of people willing to participate in telephone surveys has declined sharply, and the response rate to telephone surveys now hovers near 3%. By contrast, more people are willing to participate in surveys that are mostly conducted online or on paper like the RLS; the response rate to the new RLS is 20%. This is yet another factor that complicates drawing comparisons between the new RLS and the previous studies.]<\/p>\n\n\n\n But not all<\/em> questions are equally subject to these \u201cmode effects.\u201d While testing indicates that some questions are not comparable across different modes, others can be safely compared, and still others fall somewhere in between \u2013 the data can provide basic information about the direction of trends, even if varying modes don\u2019t produce identical results.<\/p>\n\n\n\n To help us determine which results from the new survey are comparable with those of past surveys, we conducted a telephone bridge study using a methodological approach similar to the one we used for the 2007 and 2014 landscape studies. In the bridge study, we used a random-digit-dialing method and conducted telephone interviews with 1,519 respondents separately<\/em> from the main survey. We asked these respondents the same questions that were administered to respondents in the main survey. But the 1,519 respondents in the telephone bridge study are not<\/em> included in the main RLS survey; the bridge study was conducted only as a test, exclusively for the purpose of helping us determine which questions in the new survey can safely be compared with the 2007 and 2014 results, which questions clearly cannot<\/em> be compared, and which fall somewhere in between. (Read about the bridge study\u2019s Methodology<\/a>.)<\/p>\n\n\n\n We compared the results of the telephone bridge study with those of the main address-based sample. For each question, we examined whether the balance of opinion was similar across the differing modes (phone for the bridge study, mostly online\/paper for the main survey). We also considered whether the different survey modes led to different distributions on particular<\/em> response options. For example, people who completed the survey by phone in the bridge study sometimes volunteered \u201cI don\u2019t know\u201d as a response, an option that is not available to respondents on the web and mail surveys. This sometimes resulted in fewer phone respondents choosing one of the substantive<\/em> response options, compared with those who completed the survey via other modes.<\/p>\n\n\n\n The bridge study is one important tool we used to help determine how to compare the results from the 2023-24 RLS with the results from the previous landscape studies, but it is not the only tool we used for this purpose. In addition to the bridge study, we also used other Pew Research Center analyses and experiences to help inform our decisions about whether to make comparisons between the new survey and previous RLS results. For example, we conducted an analysis in 2021<\/a> assessing whether it is possible to compare phone results with web results on key questions about religious identity, frequency of prayer, religious service attendance, and religion\u2019s importance in people\u2019s lives. Several RLS questions also have been asked on other recent Center surveys, which can provide additional information about comparability. <\/p>\n\n\n\n We drew on all this information \u2013 the telephone bridge study, previous efforts to understand mode effects, and ongoing Pew Research Center surveys \u2013 to categorize RLS questions into three buckets, color coded as green (\u201cgo\u201d), yellow (\u201ccaution\u201d) and red (\u201cstop\u201d).[102. Some questions in the 2023-24 RLS are new and were not included in either the 2007 or 2014 RLS questionnaires. This analysis focuses exclusively on questions in the new survey that were also asked (in similar or identical form) in one or both of the previous surveys. New questions were not assigned a color.]<\/p>\n\n\n\n Green questions produce results in the new survey that can safely be compared with the 2007 and 2014 RLS findings.<\/strong> There is no evidence that these questions are subject to significant mode effects. In polling lingo, they are \u201ctrendable\u201d even though the new RLS was conducted using different modes than the previous studies.<\/p>\n\n\n\n Red questions cannot<\/em> be compared with the previous studies.<\/strong> The mode effects on these questions are relatively large. Comparing the results of these questions from the new study with previous studies could be misleading. Doing so might suggest that big changes have occurred in public opinion over the last 10 or 15 years when, in reality, the different results could be produced by the mode switch alone.<\/p>\n\n\n\n Yellow questions fall between the safely trendable green questions, on the one hand, and the clearly non-trendable red questions, on the other.<\/strong> Yellow questions appear to be subject to modest mode effects, but they nevertheless can be cautiously<\/em> compared with previous results while bearing in mind the size and direction of the mode effects on the particular question at hand. <\/p>\n\n\n\n Many key questions in the new RLS can be safely compared with the 2007 and 2014 landscape studies, including the study\u2019s core questions about religious identity (also sometimes called religious affiliation, preference or adherence). The bridge study finds little evidence that people answer questions differently on the phone than online\/on paper when asked about their current religious identity, their childhood religious identity, or the religious identity of their spouse\/partner.<\/p>\n\n\n\n The bridge study yielded somewhat fewer Catholic respondents than the main survey. And among respondents who are married or living with a partner, the bridge study finds a slightly smaller share who say their spouse\/partner is religiously unaffiliated. Otherwise, though, the religious composition of respondents in the bridge study was quite similar to the religious composition of respondents in the main survey. Additionally, prior Center analysis found that questions about religious identity can be safely compared<\/a> between interviewer-administered and self-administered surveys (assuming there are no other differences).<\/p>\n\n\n\n The new survey\u2019s questions about prayer frequency can also be safely compared with the prior studies. The percentage of people who say they pray daily is somewhat higher in the telephone bridge study than in the main survey (by 3 percentage points), and the share of people who say they seldom or never pray is somewhat lower in the bridge study than in the main survey (by 4 points). However, previous Center analysis found little difference in answers to this question between a telephone survey and self-administered surveys. And both the main survey and the companion bridge study suggest that the share of Americans who pray daily has declined since 2014.<\/p>\n\n\n\n Several questions in the RLS ask respondents about personal characteristics, like their age and race. Many of these demographic variables are used to \u201cweight\u201d the survey data to known parameters derived from the U.S. Census Bureau, to make sure that the survey\u2019s sample is representative of the country\u2019s broader population. These kinds of demographic traits are mostly safe to compare with the previous RLS findings.<\/p>\n\n\n\n Here is the complete list of \u201cgreen\u201d questions that can safely be compared between the new RLS and the previous surveys (refer to the Topline for full results):<\/strong><\/p>\n\n\n\n There are several questions in the 2023-24 RLS that we cannot compare with our previous landscape studies, even though the wording is the same.<\/p>\n\n\n\n When respondents complete surveys by directly talking to another person, they sometimes give more socially desirable responses or avoid potentially controversial responses. This is called \u201csocial desirability bias.\u201d The results of the bridge study suggest that this bias may have played a role in responses to some questions.<\/p>\n\n\n\n For example, on a question about whether the country\u2019s growing immigrant population has been a change for the better, a change for the worse, or hasn\u2019t made much difference, the share of U.S. adults who say \u201cchange for the worse\u201d is 15 points lower in the telephone bridge study than on the main survey. It seems that completing the survey by talking to another person over the phone may lead some respondents to avoid giving a negative opinion about the growing immigrant population.<\/p>\n\n\n\n The bridge study and previous Pew Research Center analyses also show that the share of people who say they often attend religious services is higher in telephone surveys than in self-administered surveys. As a consequence, the results of our 2023-24 RLS question about frequency of religious attendance cannot be compared with those of the 2007 and 2014 studies.<\/p>\n\n\n\n On other questions, we cannot compare results because the percentages who say they \u201cdon\u2019t know\u201d in the phone-based bridge study are significantly higher than in the main study. For example, on a question about political party identification, the share of respondents who say they don\u2019t know or who don\u2019t answer the question is 5 points higher in the bridge study. And in a follow-up question asking those who do not identify as Republicans or Democrats whether they lean<\/em> Republican or Democratic, the share who say they don\u2019t know or who decline to answer is also higher in the bridge study. Ultimately, the share of respondents who don\u2019t identify with or lean toward either political party is 14 points higher in the bridge study than in the main survey.<\/p>\n\n\n\n Some questions are subject to potential \u201ccontext effects,\u201d suggesting they may not be comparable with previous instances in which the same question appeared in a different context. For example, the distribution of responses to the 2023-24 survey\u2019s question about the size of the government (asking whether respondents would prefer a bigger government that provides more services or a smaller government that provides fewer services) is very different in the new survey (which finds more support for a bigger government than for a smaller one) than in other Center surveys (which find a relatively even split). In the new RLS, responses to this question might have been influenced by asking it immediately after a question on whether the government provides too little, too much, or about the right amount of assistance to people in need. In the prior RLS, this question was asked in a more neutral context, and so we have decided not to compare the new findings with the old findings.<\/p>\n\n\n\n Finally, we cannot directly compare results about income levels. On all of the surveys, we ask respondents about their total household income. But because our analysis does not adjust for inflation, the value of a $40,000-$50,000 household income, for example, is different today than it was in 2007 or 2014.<\/p>\n\n\n\n Here is a list of the RLS questions we have coded as \u201cred,\u201d meaning they are not<\/em> comparable with the earlier studies, even though those earlier studies included similar or identical questions<\/strong> (in the survey\u2019s Topline, we show only the 2023-24 results on red questions and not the prior results, since they are not trendable):<\/p>\n\n\n\n Finally, there are some questions that produce results that can be compared with previous surveys, as long as the comparisons are done cautiously.<\/p>\n\n\n\n One category includes \u201cbalanced choice\u201d questions in which the main study and the phone-based bridge study get similar results for one of the substantive response options (typically the affirmative option), but slightly different results on the other substantive response option, often due to larger shares of \u201cdon\u2019t know\u201d responses and refusals to answer the question on the phone. For example, the difference between those who say \u201chomosexuality should be accepted by society\u201d in the main study and the bridge study is less than 1 point. However, the share of respondents who say they don\u2019t know or who don\u2019t answer the question in the bridge study is 7 points higher than in the main survey. And the share who say \u201chomosexuality should be discouraged by society\u201d is 7 points lower<\/em> in the bridge study than in the main survey. In other words, the share of respondents who say \u201chomosexuality should be accepted by society\u201d can<\/em> be safely compared with the prior studies, but the share who say it \u201cshould be discouraged\u201d cannot.<\/p>\n\n\n\n Other questions that fall in this category (having similar results on one substantive response option) ask about belief in heaven, belief in hell, and views of churches and religious organizations.<\/p>\n\n\n\n On some questions, responses differ in the degree<\/em> to which they affirm a response, but the total share who affirm the response, regardless of degree, is similar. For example, on a question about the legality of abortion, the share who say it should be \u201clegal in all cases\u201d is 3 points lower in the bridge study, but the share who say it should be \u201clegal in most cases\u201d is 5 points higher<\/em> in the bridge study. Taken together, the total share of respondents who say abortion should be legal in all or<\/em> most cases is just 2 points higher in the bridge study than in the main survey. In our analyses comparing the new survey\u2019s findings with past surveys, we focus on this net share \u2013 those who say abortion should be legal all or most of the time.<\/p>\n\n\n\n Here is a table that lists the questions we have deemed \u201cyellow\u201d \u2013 cautiously trendable \u2013 with a brief explanation of some points readers should keep in mind when interpreting the trends on these questions.<\/strong> In the survey Topline, these questions are all clearly labeled as yellow, and the 2007 and 2014 results are described as \u201ctrends for comparison.\u201d<\/p>\n\n\n\nUsing a bridge study and other data to decide how to compare results<\/h4>\n\n\n\n
Green: Questions we can compare with previous Religious Landscape Studies<\/h3>\n\n\n\n
\n
Red: Questions we cannot compare with previous Religious Landscape Studies<\/h3>\n\n\n\n
\n
Yellow: Questions we can cautiously compare with previous Religious Landscape Studies<\/h3>\n\n\n\n
\n\n\n\n\u2018Yellow\u2019 trend questions<\/h2>\n\n\t \n
![\"Table](\"https:\/\/alpha.pewresearch.org\/pewresearch-org\/wp-content\/uploads\/sites\/20\/2025\/02\/PR_2025.02.26_religious-landscape-study_A-01.png?w=840\")
<\/a><\/figure>\n\n\n\n![\"Table](\"https:\/\/alpha.pewresearch.org\/pewresearch-org\/wp-content\/uploads\/sites\/20\/2025\/02\/PR_2025.02.26_religious-landscape-study_A-02.png?w=1011\")
<\/a><\/figure>\n","protected":false},"excerpt":{"rendered":"