Elected representatives rely on public-facing communication to enhance their standing in the eyes of constituents and potential donors. Thanks to expanded public adoption of the internet and social media, legislators now have more means of written communication available to them than ever before (e.g., Cormack, Reference Cormack2016; Straus, Reference Straus2018; Geiger, Reference Geiger2019). Scholars have explored the implications of expanded communication options for legislative “homestyles” both in the USA (e.g., Glassman et al., Reference Glassman, Straus and Shogan2010; Grimmer, Reference Grimmer2013; Grimmer et al., Reference Grimmer, Westwood and Messing2014; Hemphill et al., Reference Hemphill, Russell and MSchöpke-Gonzalez2020) and beyond (e.g., Obholzer and Daniel, Reference Obholzer and Daniel2016; Scherpereel et al., Reference Scherpereel, Wohlgemuth and Lievens2018). Others have leveraged medium type to measure partisanship, ideology, and party brands (e.g., Barberá, Reference Barberá2015; Jones et al., Reference Jones, Noorbaloochi, Jost, Bonneau, Nagler and Tucker2018; Green et al., Reference Green, Edgerton, Naftel, Shoub and Cranmer2020). These studies share a common feature: they draw conclusions about Congressional communication using one or two platforms.
The field still lacks a baseline understanding of whether and how legislative communication varies across platforms. As a result, it is unclear whether findings based on one- or two-platform studies can be generalized beyond these platforms.
In this note, we offer the first multi-medium comparative analysis of legislative communication. To do so, we analyze US House members’ written, public-facing communications across four platforms during three legislative sessions, the 114th, 115th, and 116th Congresses (2015–2020). Using our unique data set of the official press releases, e-newsletters, Facebook posts, and tweets on Twitter issued in this period, we show that use patterns vary across platforms based on members’ institutional status, individual characteristics, and district demographics. These findings establish a baseline understanding of cross-medium communication on the most widely adopted platforms that can inform scholars of legislative communication when selecting a medium for different areas of interest.
1. Factors informing communication
Previous studies link several factors with both how frequently legislators communicate with the public, and the platforms they use to do so. These explanatory mechanisms fall into three (often overlapping) categories: institutional status, individual characteristics, and constituency characteristics.Footnote 1
Institutional status encompasses a legislator's formal and informal position within the chamber, which in turn affects access to the legislative agenda and the media. Higher-status members (e.g., party leaders, majority party members, those in the president's party) receive more media coverage, and may have less of a need for direct communication with constituents (Cox and McCubbin, Reference Cox and McCubbin1993; Meinke, Reference Meinke2009; Cook, Reference Cook2010). This does not necessarily mean that higher status members communicate less, however. Party leadership positions come both with additional resources that enhance leaders’ abilities to communicate (e.g., more staff), and with additional responsibilities (e.g., maintaining and perpetuating the party's brand) that enhance leaders’ mandate to communicate (Cox and McCubbin, Reference Cox and McCubbin1993; Heberlig et al., Reference Heberlig, Hetherington and Larson2006; Oleszek, Reference Oleszek2006; Meinke, Reference Meinke2009; Glassman, Reference Glassman2012).
In contrast, lower-status members (e.g., minority party members, members of the presidential out-party, or more extreme members) lack influence over the legislative agenda. They also receive less “free” media attention. For these members’, direct communication with the public is one of the only ways to gain the attention of constituents’ and donors (Gulati, Reference Gulati2004; Lassen and Brown, Reference Lassen and Brown2011; Straus et al., Reference Straus, Williams, Shogan and Glassman2016; Cormack, Reference Cormack2018; Russell, Reference Russell2020).
More broadly, a legislator's party affiliation likely plays some role in cultivating communication patterns. From the Democratic Message Board to the Republican Theme Team, Congressional parties have established mechanisms for coordinating communication in a way that augments the party's brand, enhancing the ability of members of the party to rapidly and repeatedly communicate with the public (Harris, Reference Harris2005; Meinke, Reference Meinke2014). These coordination efforts may translate into different medium use patterns between the two parties.
Other research posits a relationship between legislators’ individual characteristics and their communication styles. For example, more senior members are less likely to adopt social media, either because of less experience with newer technologies or because their status as incumbents has not required adoption of social media in reelection campaigns (Shogan, Reference Shogan2010). Conversely, newer members might be more comfortable with social media, having used it recently for their campaigns. These newer members might communicate less than their senior counterparts on platforms with higher start-up costs, such as e-newsletters (Cook, Reference Cook2010).Footnote 2
Other individual characteristics, including electoral vulnerability, gender, and race, are associated with different communication strategies. Members from safe districts tend to advertise issue positions on Facebook, while members from marginal districts tend to use Facebook to address district-specific concerns (Schraufnagel et al., Reference Schraufnagel, Li and Schuster2017). Research on women's communication patterns using both Twitter and e-newsletters show that women communicate more prolifically than men on these platforms—a trend that might extend to other platforms as well (Cormack, Reference Cormack2016; Wagner et al., Reference Wagner, Gainous and Holman2017). Some research suggests that legislators from ethnic groups that are under-represented in Congress might communicate more using Twitter, but less using other platforms (Ardoin, Reference Ardoin2013; Tillery, Reference Tillery2019).
Finally, district-level characteristics are known to shape a member's digital home-style (Meinke, Reference Meinke2009; Butler et al., Reference Butler, Karpowitz and Pope2012; Grimmer, Reference Grimmer2013).Footnote 3 In particular, the medium a member chooses for constituent communication is a central part of building and maintaining relationships with a member's constituency. In early stages of internet adoption, members representing higher-income districts were more likely to communicate online (Adler et al., Reference Adler, Gent and Overmeyer1998). We might continue to expect greater adoption of online platforms by members who represent districts with higher median income, education levels, and population density (Lin et al., Reference Lin, Kennedy and Lazer2017; Barthel et al., Reference Barthel, Grieco and Shearer2019; Pew Research Center, 2021). Members representing a predominately Black constituency might favor Twitter, given that Black users make up a disproportionately large number of Twitter users (Smith, Reference Smith2010; Hargittai and Litt, Reference Hargittai and Litt2011). In contrast, members representing a predominately Latinx constituency might be less likely to communicate online due to the persistent “digital divide” between white and Latinx Americans.Footnote 4
2. Data on public communication
We analyze a novel data set of four types of public-facing communication from official (e.g., not from campaign or personal accounts) press releases, e-newsletters, tweets, and Facebook posts issued by members of the House of Representatives during the 114th–116th Congresses (2015–2020).Footnote 5 These four mediums are uniquely suited to comparison for the purposes of establishing a baseline of constituent communication. First, all four feature written communication. Second, all four enjoyed nearly universal adoption in the period in question.Footnote 6 These two features distinguish the mediums we analyze from audiovisual platforms (e.g., Instagram, Snapchat, TikTok, YouTube), which were not as widely adopted between 2015 and 2020, and which would implicate sticky methodological issues about the comparison of written and non-written forms of communication.
We focus our analysis on these Congresses for two reasons. First, these three sessions provide variation in which party controls the House and/or the presidency. Second, and more practically, collection of members’ Facebook posts did not occur regularly until 2015. Because members’ Facebook accounts and posts are typically deleted when they leave office, comprehensive collection of historical Facebook posts is not possible.
Information on the rates of use come from different sources. Press release and Twitter totals are from collections archived by ProPublica, and e-newsletter totals come from DCinbox (Cormack, Reference Cormack2017). Facebook counts are from two sources: rates for the 114th and 115th Congresses are from Pew Research Center (Messing et al., Reference Messing, Kessel and Hughes2017), and rates for the 116th Congress were collected by these authors using CrowdTangle. A comprehensive discussion of data collection and data set construction is provided in the online appendix.
2.1 Summary of variables
We summarize the average use rate for each medium across all three Congressional sessions in Table 1.Footnote 7 As this table shows, rates of use differ widely by medium, although most legislators used each medium at least once in the period under examination. Legislators issue e-newsletters at the lowest rate, followed by press releases. In contrast, members use Twitter and Facebook more frequently. Intuitively, this makes sense: the first two are longer format mediums and cannot be edited once released, while the last two are comparatively short and can be edited or removed after posting.
Average yearly use rates by medium and member for the 114th, 115th, and 116th congresses)
Our key dependent variables are use rates of each of these four mediums. For purposes of cross-medium comparison, we standardize and normalize the rate of use per year for the four communication types. In the regression models that follow, we analyze pooled data across the three Congresses in order to capture the institutional variation that occurs during this period (e.g., changes in party control of the chamber).
We also include several independent variables to account for the factors that might influence legislators’ communication choices. Variables relating to institutional status include: membership in the majority party, membership in the same party as the president, position in party leadership, chair of a committee or subcommittee, party membership, and ideological extremism (distance from median DW-NOMINATE score in chamber). To account for individual factors we account for: first-term status, seniority (years served), state legislature professionalism, gender, race, and the member's district vote proportion in their most recent election. Finally, we control for district characteristics, including: median income (in $100,000s), population density (logged), and indicators for majority-minority districts (Black or Latinx),Footnote 8 and the proportion of the population that has completed high school. Details on these variables and their distributions are given in the online appendix.
3. Results and implications
We assess the link between institutional, legislator, and district-level factors on average use of all four types of Congressional communication using ordinary least squares (OLS) regressions.Footnote 9 Data are pooled across Congressional sessions, and communication rates are standardized and normalized. As a result, a one unit change in an independent variable corresponds with a coefficient-sized shift in the standard deviation of the each medium's yearly use rate. All else equal, if the estimated coefficient associated with a given variable is 1, then the dependent variable is expected to increase by 1 standard deviation if that variable takes on a value of 1.
We present the regression results as a coefficient plot, with dashed vertical lines indicating zero, points indicating coefficient estimates, and light gray bars indicating the 95 percent confidence interval (Figure 1). For the full regression table, see the online appendix.
Coefficient plot for each of the four OLS regressions.
Three broad patterns emerge in these results: (1) variables with similar use rates across mediums, (2) variables with statistically significant influence on two or fewer mediums, and (3) variables with use rates that differ in both sign and significance across two or more mediums.
First, a number of variables are related to every, or almost every, medium in the same manner. Leaders (of both parties) communicate more than other legislators, which is in line with research stressing access to greater resources and responsibilities for those in higher positions. Beyond that: (1) women legislators communicate more on average than their male counterparts (aside from e-newsletters); (2) those representing majority Latinx districts communicate more on average (aside from e-newsletters); and (3) as the proportion of the district that has more than a high school degree increases, so too does the amount communications on all mediums except press releases. Additionally, there are a number of variables that exhibit a null relationship across all four regressions. These variables include: state legislative professionalism, vote proportion received in the last election, whether the legislator is Black, and whether the majority of the district is Black. These relatively stable relationships across medium contribute to our better understanding of how members communicate, and suggest situations in which research could obtain similar results regardless of the medium.
Second, we note several variables that correspond with significantly different use rates for two or fewer mediums. Three variables relate to significantly different use rates for only one medium: (1) Members of the majority party use Twitter more; (2) committee or subcommittee chairs send more e-newsletters; and (3) members representing higher median income districts send fewer e-newsletters. Additionally, two variables correspond with significantly different use rates in the same direction for two mediums. Members of the president's party use both Twitter and Facebook at lower rates, and Latinx members utilize both press releases and Facebook at lower rates than other members of Congress. Researchers interested in the relationship between these variables and Congressional communication should take care to distinguish between medium-specific impacts and general communication styles.
Finally, and perhaps most importantly, there are some variables that see opposing, statistically significant relationships across multiple mediums. Three of these variables are associated with substantively large differences in medium use: (1) Democratic members are more likely to communicate via both press releases and Twitter, but less likely to send e-newsletters; (2) more ideologically extreme members are more likely to use Twitter, but less likely to use both press releases and e-newsletters; and (3) first-term members are more likely to communicate using both press releases and Facebook, but are less likely to send e-newsletters. Two additional variables, seniority and district population density, show opposing relationships that are substantively smaller. More senior members are slightly more likely to issue press releases and slightly less likely to use Facebook, and members in more populated districts are slightly more likely to use both e-newsletters and Twitter, but are slightly less likely to issue press releases. These findings suggest that researchers should take the greatest care in medium choice when examining these variables, as their results will differ substantially based on which medium they analyze.
4. Discussion and conclusion
These cross-medium analyses provide a baseline understanding of how differently situated legislators communicate with the public. Our findings indicate that medium choice matters, as who uses what—and how often—systematically varies across mediums. For example, a researcher turning to e-newsletters may come to different, opposing conclusions concerning use patterns for Democrats and Republicans or the engagement style of ideological extremists than someone turning to Twitter for data.
These implications do not mean that researchers should avoid using certain platforms in their analysis or that they must use all platforms in each study. Rather, we provide a framework within which to justify medium choice, and an indication for how studies may (or may not) be biased in their estimates and conclusions.
In general, our results illustrate important links between institutional status, legislator and district characteristics and platform use. Some of these relationships (or lack thereof) seem stable across mediums. For example, party leadership is related to greater communication across all mediums.
In addition, we highlight relationships that differ across mediums. These include the role of party, majority party status, and ideological extremism, which are often the focus of scholars studying communication, elections, policy, and polarization. As research continues on how legislators communicate with the public, leveraging public communication to measure party brand and ideology, and develop a better understanding of when and why elite rhetoric shapes public opinion and behavior, this study provides insights on the conditional nature of medium use.
What we offer here is an empirical baseline of how members use four of the most widely adopted, written forms of constituent communication. This is an essential first step in deepening our understanding of legislative communication, and one that we hope will inspire studies that draw out the theoretical implications of these findings, examine the relationship between medium and home-styles, delve into the content and priorities expressed on these mediums, and expand this comparison to include audiovisual mediums as representational styles continue to evolve.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/psrm.2022.28. To obtain replication material for this article, please visit https://doi.org/10.7910/DVN/XRZ1UJ.
Open access
