RELIGIOUS LANGUAGE AND POLITICIANS

It is nothing new for politicians to use religious language when communicating with constituent groups. Even with a formal separation of church and state, the American people have generally “not denied the political realm a religious dimension” which is “expressed in a set of beliefs, symbols, and rituals” often referred to as American Civil Religion (Bellah Reference Bellah1967, 3–4). Religious appeals are not only tolerated but fairly common because they work. Chapp (Reference Chapp2012, 4) argues that “religious rhetoric gains its unique political command because it is well equipped to resonate with individuals’ emotions and identities—two factors that, not coincidentally, are central to political persuasion.”

Both Bellah (Reference Bellah1967) and Domke and Coe (Reference Domke and Coe2008) highlight the common use of religious rhetoric by U.S. presidents over time. However, Domke and Coe (Reference Domke and Coe2008) point to Ronald Reagan's presidential nomination in 1980 as a turning point in religiopolitical communication in the United States.Footnote ¹ While politicians used religious communication pre-1980, the authors argue that political leaders have since used a God Strategy, where religious language is “carefully crafted” and “employed by politicians to connect with religious inclined voters” (p. 8).

The use of religious language acts like a heuristic, signaling a set of beliefs and values encouraging voters to consider religious identities when making political choices (Abelson and Levi Reference Abelson, Levi, Lindzey and Aronson1985; Calfano and Djupe Reference Calfano and Djupe2009). However, “sometimes these religious signals are intended for the eyes and ears of all Americans, and other times, they are implemented in targeted ways, as veritable ‘dog whistles’ that only distinct segments of the population fully receive” (Domke and Coe Reference Domke and Coe2008, 8). Calfano and Djupe (Reference Calfano and Djupe2009) discuss how this “code” can be used stealthily to appeal to in-groups without out-groups noticing.

Domke and Coe (Reference Domke and Coe2008) systematically track religious language used in presidential communication over time, finding that Presidents use religious code strategically. The communication may be sincere, but it is not random. They show that presidents have used the code over many decades, but religious communication increased decisively after Reagan's first election. Perhaps surprisingly, given the religious divide among American partisans since the 1980s (Layman Reference Layman2001), the authors also find that presidents of both parties use religious signals, targeting Christian fundamentalists, conservative evangelicals, and conservative Catholics (p. 19). Domke and Coe (Reference Domke and Coe2008) rule out alternative explanations for why and when presidents use the religious code. They find no evidence that presidents use it only when facing an election or during war time. As stated above, it cannot be explained completely by party affiliation of the president.

Building on this work, Calfano and Djupe (Reference Calfano and Djupe2009) question whether the religious code works and ask which people are paying attention. Does carefully scripted religious language by presidents and political elites, more broadly, ignite religious or non-religious identities among constituents? They design an experiment testing Republican religious messages targeting white evangelicalsFootnote ² and find that the strategy works. White evangelicals indicate their likely support for candidates based on their use of the religious code and associate these candidates with the Republican Party. Yet, for the most part, the religious code did not affect Catholic and mainline Protestant participants' candidate evaluations. Finally, Calfano and Djupe (Reference Calfano and Djupe2009) also examine racial factors, showing that African American candidates benefitted from using the code as well.

To sum up the above scholarship, there is evidence that presidents and political elites use religious language strategically, especially since the 1980s. These appeals likely influence perceptions of the religious and non-religious, Republicans and Democrats, and likely have electoral consequences. For example, ballot measures with a moral component have consistently brought voters to the polls (Biggers Reference Biggers2014). More broadly, the ways political elites use religious rhetoric determine the “exact role that religion plays in American elections, political culture, and the representative dynamics of the country” (Chapp Reference Chapp2012, 4).

Most research in this area analyzes religious speech by presidents, but there is much less work considering religious rhetoric by members of Congress. How do they use the religious code to connect with their varied constituencies? With the rise of social media, MOCs have the opportunity for daily messaging with their local constituents (and other interested parties) even when away from the home district.

MEMBERS OF CONGRESS USING TWITTER

Researchers have done considerable work on members of Congress' adoption and use of the Twitter platform in the last several years. Work across several disciplines probes this data to better understand and explain Twitter communication by politicians and specifically, MOCs. After the Obama campaign's extensive use of Twitter during the 2008 presidential campaign, candidates increasingly use the platform to raise money and communicate their issue positions (Tumasjan et al. Reference Tumasjan, Sprenger, Sandner and Welpe2010; Evans, Cordova, and Sipole Reference Evans, Cordova and Sipole2014; Evans, Brown, and Wimberly Reference Evans, Brown and Wimberly2018). Elected officials use Twitter as part of a broader media strategy where they want to gain the attention of media outlets, depending on journalists to relay the content of their tweets as they would official press statements (Lieber and Golan Reference Lieber, Golan, Stromback and Kiousis2011; Wallsten Reference Wallsten and Solo2014; Bernhard, Dohle, and Vowe Reference Bernhard, Dohle and Vowe2016; Shapiro and Hemphill Reference Shapiro and Hemphill2017).

Evans, Cordova, and Sipole (Reference Evans, Cordova and Sipole2014) are among some of the first researchers to examine how politicians use the social networking site to communicate with constituents. In their study, tweets become data to understand how candidates for the United States House of Representatives communicate with the media and constituents during their 2012 campaigns. The authors find that candidates tweeted about their personal lives approximately one-third of their time on Twitter (and about issues 11% of the time). Once elected, not every member of Congress tweets the same. There are varying rates of participation and styles (Evans, Cordova, and Sipole Reference Evans, Cordova and Sipole2014). Additionally, the content of MOCs' tweets changes depending on whether they are in recess or in session. Glassman, Straus, and Shogan (Reference Glassman, Straus and Shogan2010) show that session tweets tend to be policy focused while recess tweets are district focused.

The earlier work collecting this data notes that the conclusions will change as more members use Twitter. Its use has increased drastically over the last several years. The 112^th Congress may be considered the first “Tweeting” Congress with 95% of all members on Twitter (Hemphill, Otterbacher, and Shapiro Reference Hemphill, Otterbacher and Shapiro2013; Gervais and Morris Reference Gervais and Morris2014). More recently, nearly every MOC in the 115^th Congress had a verified Twitter account, so it is easier than ever to gather data on the content of these communications. Twitter is an ideal platform for exploring MOCs' strategic use of religious code because most members use it, and because it allows for the emphasis of both the personal and shared aspects of religious identity.

MOCs may use religious code to communicate personal details that signal a shared identity with targeted constituents. McGregor's (Reference McGregor2017) research shows that personalized tweets cause voters “to manifest positive feelings of intimacy towards candidates” and “create impressions of closeness to and imagined relationships with the candidate, which increases the likelihood of electoral support” (p. 2). This personalization tends to work better for male candidates regardless of party. Tweets by MOCs are not random or haphazard but designed to carefully manage an image in a world increasingly connected by social media (Leary and Kowalski Reference Leary and Kowalski1990; McGraw Reference McGraw, Sears, Huddy and Jervis2003; Meeks Reference Meeks2016).

The need for impression management may be greater when “people encounter moments of acute public awareness or intense scrutiny” (Meeks Reference Meeks2016, 1052). MOCs have used Twitter to respond to events tied to the Black Lives Matter Movement. Dancey and Masand (2019) write that it can be especially important for certain members to speak to a diverse public during times of tragedy, protests, and court decisions. In this way, MOCs move beyond district-based representation and the electoral connection to something broader, descriptive representation. In the same way, MOCs may use their religious identities to not only manage their identity when relevant for fulfilling a goal, like getting elected, but also as descriptive representation and part of a larger God Strategy (Leary and Kowalski Reference Leary and Kowalski1990; Meeks Reference Meeks2016). It is also likely that religiously toned tweets are tied to moments of crisis or events of great magnitude in the public square.

The relationship between religion and politics in the United States is important but can be difficult to disentangle. While there is not a lot of evidence that religious leaders use Twitter to signal political identities (Burge and Williams, Reference Burge and Williams2019), we know that politicians use religious language as part of a God Strategy, and they likely use that strategy on Twitter. We want to understand how, when, and which MOCs use the religious code on Twitter.

EXPECTATIONS FOR MEMBERS OF CONGRESS USING RELIGIOUS LANGUAGE ON TWITTER

Building on this past scholarship, we have a few expectations for how, when, and which MOCs tweet religious language based on party identification and gender. Because of the broad religious landscape and the winner-take-all nature of many U.S. elections, “candidates cannot afford to ignore religion, nor can they afford to privilege an particular faith tradition” (Chapp Reference Chapp2012, 6). However, partisanship frames and shapes almost every aspect of current congressional politics, and it likely influences religious tweeting by members of Congress (Evans, Cordova, and Sipole Reference Evans, Cordova and Sipole2014; Meeks Reference Meeks2016). We expect to find more religious tweeting from Republicans because they are generally “more inclined and better positioned to capitalize on a convergence of religion and politics” (Domke and Coe Reference Domke and Coe2008, 9). However, Democratic presidents have used religious language in the past and made inroads after the 2004 elections (Domke and Coe Reference Domke and Coe2008). Yet, Democratic politicians risk alienating a growing number of voters who do not affiliate with any religious group but tend to vote Democrat. Even civil religion appeals can leave non-Christians and the non-religious feeling unrepresented and excluded (Chapp Reference Chapp2012). Weighing all of these considerations in mind, we hypothesize that:

The Republican Party has been closely linked with the Christian Right and its voters since the 1980s, so we expect that Republican MOCs will tweet not only more often but also with more variation, focusing on Judeo-Christian terms. For instance, Domke and Coe (Reference Domke and Coe2008) note the increase in references to Jesus (twice as many mentions as God) in presidential Christians communications after 1980. They write that “for most moderates and nonbelievers, this is a superfluous distinction,” one that is lost on them but “for the Christian faithful, it's the ultimate narrowcast message” (p. 97). So, we hypothesize that:

Previous studies show that female MOCs and candidates use Twitter distinctly from their male counterparts, as discussed in the above section. However, much less is known about whether the representative's gender correlates with distinct religious tweeting or messaging. Sterk (Reference Sterk2010) wrote a 10-year review on “Faith, Feminism and Scholarship in The Journalism of Communication and Religion,” noting that the “intersection of gender, communication and religion” remains undeveloped (p. 207).

Survey research demonstrates repeatedly that females personally identify as more religious than males (Gilligan Reference Gilligan1982; Cornwall Reference Cornwall, Bahr and Peterson1989; Mullikin Reference Mullikin2006; Putnam and Campbell Reference Putnam and Campbell2010), and they are more interested in religion and attend worship services more often (Yinger Reference Yinger1970; Cornwall Reference Cornwall, Bahr and Peterson1989; Batson, Schoenrade, and Ventis Reference Batson, Schoenrade and Ventis1993; Mullikin Reference Mullikin2006) across various religious traditions, so we might expect female MOCs to use the religious code more often than male MOCs to personalize themselves (Meeks Reference Meeks2016). However, Niven and Zilber (Reference Niven and Zilber2001) show that congressional websites for female leaders have slightly fewer details about their families compared with male MOCs and strongly assert their aptitude for their work, devoting “more space than men to personal qualifications (such as their education and previous experience) that have some relevance to the job” (Niven and Zilber Reference Niven and Zilber2001, 399). Male and female candidates and MOCs are trying to communicate similar messages to voters (Zilber and Niven Reference Zilber and Niven2000; Panagopoulos Reference Panagopoulos2004)—perhaps because of the media's bend toward stereotypes. We know that in the past, media have traditionally relied on long-standing gender stereotypes to cover candidates for political office (Kittilson and Fridkin Reference Kittilson and Fridkin2008). Abbott (Reference Abbott2006) shows how media portray males with religious messaging as sensitive and caring, familial and warm, while self-identified feminists were often framed as more masculine.

Given this work on gender stereotypes, we think the more likely scenario is that female MOCs will tweet religious messages less than their male counterparts. This may be due to their out-group status in a still male-dominated field (Evans and Clark Reference Evans and Clark2016), or males may use religious language more often because they tend to benefit from the resulting association (Abbott Reference Abbott2006; McGregor Reference McGregor2017) where women may not. We hypothesize that:

Finally, we also anticipate that the timing of religious code is not random. In his examination of civil religion rhetoric in speeches by presidents, Bellah (Reference Bellah1967, 2) suggests that references to God are almost always used “on solemn occasions.” MOCs also use religious code during particular seasons and in response to certain events. Domke and Coe (Reference Domke and Coe2008) examine Christmas communications overtime and find that references to “Christ” as opposed to “God” have become more common more in recent presidential Christmas communications. Politicians' use of the phrase “war on Christmas” has also gained traction in recent election cycles (Claassen Reference Claassen2015). We should notice these patterns among MOCs on Twitter as well. Our final hypothesis is:

We code for religious words and phrases based on an established set of religious language, previously examined among presidential communication (Domke and Coe Reference Domke and Coe2008; Hughes Reference Hughes2019).

DATA AND METHOD

Acquiring a significant database of the tweets composed by members of Congress is a logistically difficult task. The first hurdle that must be cleared is the acquisition of the Twitter usernames of all present members of Congress. Unfortunately, no government entity has officially been given this task, therefore other means needed to be pursued to compile a database. In 2009, Twitter introduced a feature called “lists” which allowed a user to create a custom timeline of user accounts that were grouped around a specific purpose (Stone Reference Stone2009). For instance, the New York Times Twitter account has a list of all its reporters' official Twitter accounts (https://twitter.com/nytimes/lists/nyt-journalists). C-SPAN's Twitter account has a number of helpful lists, including collections of U.S. Representatives (https://twitter.com/cspan/lists/u-s-representatives) and Senators' Twitter accounts (https://twitter.com/cspan/lists/senators). These lists were used as the basis to collect the necessary tweets. However, this is not a perfect solution, as C-SPAN's lists contain accounts that are not tied to a specific member of Congress. For instance, their Senate list contains the account of the Senate Indian Affairs Committee. Accounts like these were eliminated from the analysis. In total, 452 accounts for U.S. Representatives were scraped, along with 103 U.S. Senators.Footnote ³

After the list of accounts was acquired, it was necessary to scrape the tweets from each of these accounts. Twitter offers an application programming interface (API) to the general public; however, there are a wide variety of APIs that constitute tradeoffs to a researcher. The API that is most widely employed by researches is the REST API, which is free to use. However, a particular downside is that it can only scrape the previous 3,200 tweets from a user's timeline.Footnote ⁴ Because of the vast differences in the frequency of tweets from individual members of Congress, reaching the 3,200 tweet threshold could result in acquiring 30 days or 3 years of tweets. In most cases, this dataset contains the maximum allowable number of tweets. For instance, of the 103 Senator accounts that were scraped, 85 of them contained at least 3,000 tweets indicating that the user's entire timeline had not been acquired through this data collection. This could lead to some methodological issues when making conclusions about social media chatter when looking at Twitter activity several years ago. For instance, one of the most influential Congressional Twitter accounts in 2010 belonged to Michelle Bachmann (Carr Reference Carr2010), who decided to not run for reelection in 2012 and therefore is not in the dataset (Allen Reference Allen2013). Other accounts like Senator John McCain's has been quite active with a total of 14,400 tweets, but because of the limitation of the API only the last 3,200 were collected. As a way to account for the unreliability of this API, the dataset was truncated to include just tweets that were sent from January 1, 2017 until April 4, 2018. In the smaller dataset, all but two of the total Twitter accounts appear in both years. The scraping process was conducted on April 4, 2018 using the rtweet package written for the R statistical software program (Kearney Reference Kearney2018). In total, 1,502,231 tweets were collected from members of Congress, with 674,785 being sent in 2017 and 2018.

FINDINGS

Volume versus Concentration

One of the primary methodological problems when dealing with the usage of religious language on social media is coming to a conclusion about which is a better measure: volume or concentration. To measure the volume of religious language, a simple word count is sufficient. This obviously has the benefit of simplicity; however, volume is obscured by the fact the number of tweets does not correlate directly to the share of seats in the Congress. Figure 1 below displays the number of tweets sent per month by four groups: Democrat men, Democrat women, Republican men, and Republican women. Democrat men lead the way in tweet volume in most months of the data sample. An average month sees this group sending out approximately 20,000 tweets. Republican men send out just over 15,000 tweets per month, on average. Women send a lower volume of tweets, but that is more a function of how well represented they are in the halls of Congress. In fact, female Democrats use Twitter at a higher rate than their male counterparts. For instance, female Democrats tweets made up 33% of all tweets in the collection period, but only makeup 19.8% of Congress. The same is true for female Republicans, who hold 3.9% of all seats in Congress but are 8.8% of all the tweets sent in the sample.

Figure 1. Total tweet volume by month

Figure 2 displays this point even more clearly, as it displays the average number of tweets per month from January 2017 to April 2018 for each of the four major groups being discussed. The gap in tweet volume here is not based on gender, but instead is a partisan divide. For instance, a Democrat male tweets at nearly twice the rate as a Republican male, while a Democrat male tweets about 40% more than their GOP female counterparts. Taken from this angle, it appears that even though Democrats make up a smaller portion of all member accounts, their overall Twitter activity makes them much more consequential when taking to social media to share their message. As such, it seems that Democrats are “punching above their weight” on the Twittersphere, which means that they are more likely to have an impact on the discussion and framing of current political issues than the GOP.Footnote ⁵

Figure 2. Average number of tweets per party and gender

Several realities come into focus now. First, the volume of tweets by members of Congress is staggering. In March of 2018, members sent out nearly 50,000 tweets in 31 days. In such a high-velocity environment, the only way for a message to cut through the noise is for it to be repeated many times. Second, the average number of tweets sent by Democrats is much higher than Republicans in the last year. This indicates that, on balance, the Democratic members of Congress are considering Twitter to be an effective tool to reach their constituents. However, how can these findings be reconciled to create an accurate representation of religious language online? The rest of this paper will rely on the concentration approach. What this means practically is that each specific mention of a religious term will be calculated for each of the four groups under study and then divided by the total number of tweets sent out either by that group during the entire time frame or during that month of Twitter activity. The end result is that the likelihood of each group to use religious language will be clear; however, the total impact of these tweets might be obscured. For instance, if both Democrat men and Republican women used the word “faith” in 1% of tweets from their respective group that does not indicate that the average Twitter user would be just as likely to see the term used by either group, because Democrat men have twice the monthly tweet volume.Footnote ⁶ These results should be interpreted more on the intent by the members of Congress than the impact that they have on their constituents and followers (which is much more difficult to measure).

RELIGIOUS CODE

The seemingly most appropriate place to begin this assessment is by considering how often members of Congress use the most generic and ubiquitous religious word: “God.” This term is a common part of discourse in the United States and the phrase “God Bless America” has become an oft-repeated phrase to end a political speech by an elected official. Because of the nature of the term, it is well suited to gauge whether religious language (broadly defined) is utilized on Congressional Twitter accounts. The figure below displays the frequency of the word God as a percentage of total tweets sent out that month. In total, the word “God” appeared in 2,226 tweets or 0.33% of all those scraped. Of all tweets that contained “God,” just 82 contained the full phrase “God Bless America” (Figure 3).

Figure 3. Mentions of God by members of Congress

The clear result is that the word “God” appears much more frequently among Twitter accounts from Republican male politicians than any of the other three groups that were assessed. There are some months where over 0.25% of all tweets contained God, mainly at the end of 2018 and into early 2019. In an average month between January of 2017 and April of 2018, Republican males include the word “God” in their tweets in approximately one out of every 450 total tweets. The fact that members of Congress from either party have not ramped up their usage of God seems noteworthy as well as the fact that female members of Congress do not use “God” in any significant frequency compared to their male counterparts. Recall that in the last few years, female Twitter accounts produce more volume—it appears that religious language is not part of increased social media activity. Speaking more broadly, it appears that there is not any type of cyclical pattern throughout the year, as well. As such, a reasonable conclusion is that the word “God” is not necessarily anchored to a holiday or season (Figure 4).

Figure 4. Tweets sent in March of 2018 by members of Congress

March of 2018 is emblematic of the overall differences in the usage of “God.” For the month, there were 47,432 tweets sent out by members of Congress. The total share of those tweets by Republican males was 16,325 or 34.4% of the total volume. In total, the word “God” was invoked 162 times by Republicans and Democrats, but Republican males were responsible for 118 of those mentions (72.8%). On the other hand, Republican females sent out 4.9% of the total tweets in March of 2018, but their God mentions were 5.6% of the total. Democrats were much less likely to use “God.” While Democrat males made up 39.9% of the total volume, they only accounted for 12.3% of the God tweets; for female Democrats, it was 20.1% of total volume, but only 9.3% of God tweets. This pattern is replicated over a number of months, with a clear conclusion: Republican men send out the vast majority of all tweets containing the word “God,” while both female and male Democrats use the term sparingly.

Expanding the discussion to other religious words that could be potentially used by members of Congress, Figure 5 displays the number of total tweets that contained one of four religious terms divided by the total number of tweets composed by each of the four groups. The first thing that becomes apparent is that religious language is exceedingly rare by either gender of both parties. The term that appears the most, “God,” is only used by Republican men about once per 200 tweets. Most of the other terms occur less often, such as “faith,” appearing in one tweet per 500. The results from the previous analysis are carried through here: Republican men, by and large, are more comfortable with using religious language on Twitter compared to the other three groups. In addition, the results indicate that Republicans are more likely to use religious language than Democrats, no matter which term is being observed.

Figure 5. Mentions of specific religious terms

The other major conclusion is the vast differences in the frequency of the four different words. It seems appropriate that these words can be divided into two camps, generally religious terms such as “god” and “faith,” and specific Christian terms such as “Bible” and “Jesus.” For instance, Republican men were over 20 times more likely to use “God” than “Bible” and Democrat women were 30 times more likely to tweet “Faith” than “Jesus.” The clear indication here is that members of Congress are much more reluctant to use specific Christian terms, while they are less hesitant to talk about religious imagery more generally. One has to wonder if even Republican members of Congress are concerned with using inclusive language on Twitter, even though a significant portion of their constituency come from Christian traditions that would welcome the usage of the term “Jesus,” for instance.

To illustrate this point further, Figure 6 displays the number of instances in which members of Congress tweeted references to specific Bible verses since 2008. Out of 700,000 total tweets collected, just 645 contained a reference to a verse in the Bible, which is 0.01% of all tweets in the datasetFootnote ⁷. What makes this number even more interesting is that of the 645 Bible verse tweets, 320 of them came from one member: Republican John Shimkus. Beginning in 2011, Shimkus made it a point to include a tweet each day that came out of the Lutheran devotional that he was using each morning. Shimkus noted that if he weren't in Congress he would have gone to seminary, and he liked to use his platform to bear witness to his faith (Semnani Reference Semnani2011). The other frequent tweeter of Bible verses was Senator Marco Rubio, who was a fierce rival of Donald Trump during the 2016 Republican primary. Media outlets accused Rubio of using Bible verses as a way to criticize or “subtweet” Trump's action as President. Rubio would tweet verses such as Proverbs 26:11, “As dogs return to their vomit; so fools repeat their folly,” which was seen by many to be directed toward the President (Lange Reference Lange2017). No other member of Congress referenced the Bible more than 10 times, and the top 10 accounts that tweeted Bible verses were all Republicans, with nine out of 10 being male Republicans. Clearly, directly quoting the Bible happens rarely and when it does happen, it is nearly exclusively coming from Republican members of Congress.Footnote ⁸

Figure 6. Which members tweet Bible verses?

WHEN IS RELIGIOUS LANGUAGE USED BY MEMBERS OF CONGRESS?

Having described how the usage of religious language varies by party and gender, an important second question is: is religious language used spontaneously when a member feels inspiration to talk about matters of faith, or is it in response to specific events occurring in current events? Figure 7 below displays the usage of five different terms that are related both generically to religious language (faith, God, pray/praying/prayer/prayed), as well as specific to Judeo-Christianity (Bible and Jesus). The bars represent the percentage of total tweets each month that contain the word in question. This is done as a way to correct for the fact that the volume of tweets increased in 2017 and 2018.

Figure 7. The usage of different religious language over time

Looked at broadly, it is clear that most religious language does not appear at any reasonable frequency. For instance, in the average month, the word “Jesus” is used in 0.04% of the tweets by members of Congress, compared to 0.4% for God, 0.03% for Bible, 0.2% for faith, and 1% for the variations of prayer. As previously discussed, members of Congress do not use religious language at a frequency that would be perceptible to almost any of their followers. Because of this incredible infrequency of religious language, it makes it nearly impossible to determine if there are any seasonal effects for these terms being employed. As such, trying to tie most of these terms to a specific seasonal event (i.e., Christmas or Easter) would be inadvisable. The only term that comes close to appearing at an interval that is worth further exploration is the variations of the term “prayer.” As an example, the term was used 1,250 times in June of 2017, or 2.7% of all tweets that month. This frequency is 10 times greater than the usage of Bible.