Cross-Platform Partisan Positioning in Congressional Speech

• Published date: 01 September 2024
• DOI: http://doi.org/10.1177/10659129241236685
• Author: Jon Green,Kelsey Shoub,Rachel Blum,Lindsey Cormack
• Date: 01 September 2024
• Subject Matter: Articles

Article

Political Research Quarterly

2024, Vol. 77(3) 653–668

© The Author(s) 2024

Article reuse guidelines:

sagepub.com/journals-permissions

DOI: 10.1177/10659129241236685

journals.sagepub.com/home/prq

Cross-Platform Partisan Positioning in

Congressional Speech

Jon Green

1

, Kelsey Shoub

2

, Rachel Blum

3

, and

Lindsey Cormack

4



Abstract

Legislative activity—whether votes or communications—is often represented in a single partisan or ideological di-

mension. But as lawmakers communicate in various venues (e.g., traditional, direct, or social media), the extent to which

these estimates are interchangeable—ref‌lecting a common underlying dimension—is unclear. We estimat e a partisan

dimension in members’tweets, Facebook posts, e-newsletters, press releases, and one-minute House f‌loor speeches for

the 116th U.S. Congress and test the extent to which representations remain consistent across different venues. We f‌ind

that while Democrats are consistently separable from Republicans, members’relative intra-party positions frequently

shift between venues. This is likely driven by differences in the affordances and audiences present in each venue, as venues

with more nationalized audiences (such as social media) show higher levels of rhetorical polarization than venues with

more local audiences (e-newsletters). These results suggest that the level of polarization we observe depends on where

we look, and that the scholars of congressional communication should explicitly consider the input they use to measure

partisanship.

Keywords

congressional communication, text-as-data, ideology, partisanship

Members of Congress have long used communications to

stake out positions, claim credit for accomplishments, and

increase name recognition (Fenno 1978;Mayhew 2004).

The centrality of public messaging has increased in recent

decades, with congressional off‌ices devoting a greater

share of their resources both to communications staff and

legislation designed to make political statements (Lee

2016). In addition, contemporary Congressional parties

are marked by intra-party fractiousness (e.g., Blum 2020;

Clarke 2020;Gervais and Morris 2018), which incen-

tivizes members to use messaging to distinguish them-

selves from other factions within their party. Today,

members have more venues at their disposal from which

to accomplish these communication goals than ever

before.

Different venues are likely to be useful for different

forms of communication. Venues differ in their

affordances—some allow for longer-form messages,

some have character limits, some are shared publicly,

some are sent directly to those who have opted in, and so

on. The audiences in each venue likely differ as well,

although the overlap (or not) between who reads a

member’s press releases and who reads their Facebook

posts is not directly observable. Members’off‌ices might

modify their communications by venue to appeal to

different anticipated audiences, or they might create one

set of messages that can be used across every venue. We

frame these possibilities as hypotheticals, as little is

known about how members use different venues. While

previous work has demonstrated cross-member variation

in the use of venues (Blum, Cormack, and Shoub 2023),

the extent to which legislators engage in differential

partisan positioning across venues is unclear. The

1

Duke University, Durham, NC, USA

2

University of Massachusetts Amherst, Amherst, MA, USA

3

University of Oklahoma, Norman, OK, USA

4

Stevens Institute of Technology, Hoboken, NJ, USA

Corresponding Author:

Kelsey Shoub, School of Public Policy, University of Massachusetts

Amherst, Thompson Hall, 200 Hicks Way, Amherst, MA 01003, USA.

Email: kshoub@umass.edu

question of how members position themselves across

venues has important implications for how we understand

partisan positioning in Congress, polarization, and the

measurement of both quantities.

There are two plausible answers to this question. First,

there may be a single latent dimension that consistently

organizes legislators’partisan and ideological positioning

across venues (H1: One-Dimension Hypothesis)as

members face constraints on the amount of time and

staff‌ing resources they can devote to crafting distinct

messages for each venue. If members assume similarities

between each venue’s potential audiences, they might also

not see a need for individualized messaging. If this is the

case, then individual legislators’positioning in a single

partisan dimension should be similar between and within

parties across venues. Such a f‌inding would be further

evidence that American politics has indeed collapsed to a

single partisan dimension, as legislators would exist in

similar orderings across all venues. In this case, re-

searchers will obtain interchangeable estimates of latent

partisanship and/or polarization whether they analyze roll

call votes, Facebook posts, or one-minute f‌loor speeches.

Alternatively, it might be the case that how members

communicate differs based on venue. In this scenario,

Democrats would still be consistently separable from

Republicans, but within-party positioning would vary by

venue (H2: By-Venue Hypothesis). Someone with an out-

sized national presence may use more extreme rhetoric on

a medium like Twitter but communicate a more moderate

message to their constituents who receive e-newsletters.

To the extent to which this is the case, it would mean that

estimates of relative partisan positioning in congressional

communication are not interchangeable across venues.

Scholars interested in using such estimates in their

research would therefore need to consider—in both the-

oretical and practical terms—how their f‌indings may

depend on the venue in which congressional communi-

cation is observed.

To assess these competing hypotheses, we consider

f‌ive venues: social media in the form of Twitter and

Facebook, “traditional”outreach through off‌icial

e-newsletters and press releases, and institutional content

in one-minute speeches. These venues share key features:

members are free to choose the topic of discussion (unlike

in speeches given during f‌loor debates regarding specif‌ic

bills), and communications using these f‌ive venues are

aimed to some extent at audiences outside of Congress. In

a departure from previous studies, which studies con-

gressional communication by isolating a single venue, we

consider them simultaneously (but see Blum, Cormack,

and Shoub 2023;Cormack and Gulati 2023). For each

venue, we collect the text of speeches, posts, and mes-

sages and put them through rigorous data cleaning steps to

allow for comparable estimates across venues. We then

use multinomial inverse regression (Taddy 2013,2016)

with member f‌ixed effects to estimate the partisan valence

of words in each setting and use the results of this pro-

cedure to generate venue-specif‌ic projections of members

into low-dimensional space.

These models detect within-member differences across

venues, supporting the by-venue hypothesis. This mani-

fests as a Simpson’s paradox: members’estimated rhe-

torical partisanship carries relatively strong overall

correlations across venues because Democrats and Re-

publicans are generally separable from each other, but

these cross-platform correlations are considerably weaker

within each congressional caucus. This has important

theoretical and practical implications. By way of dem-

onstration, we follow-up on our main tests by showing

how choice of venue might inform estimates of chamber-

level polarization. We show that distance between party

medians and (to a lesser degree) intra-party unity vary

depending on where such quantities are measured.

In short, ours is one of the f‌irst projects to derive

multiple venue-specif‌ic scores for the same set of legis-

lators based on member-controlled communications for

the same period.

1

In doing so, we bring into conversation

work on legislative studies, political communication, and

approaches to text analysis to making both theoretical and

empirical contributions. Substantively, we echo

Tausanovitch and Warshaw(2017) in f‌inding considerable

differences in intra-party positioning between venues and

then show that this has implications for our impressions of

Congressional polarization, contributing to on-going

discussions in this area (e.g., Crespin and Rohde 2010;

Lee 2016;McCarty, Poole, and Rosenthal 2015). Em-

pirically, we engage with discussions of how text pre-

processing can inform results and for theory driven pre-

processing decisions when working with text-as-data

(Denny and Spirling 2018), providing guidance on key

questions that arise when working with legislative com-

munications. Further, these f‌indings highlight that venue

choice matters, especially for those concerned with intra-

party dynamics. Certain venues are more ref‌lective of each

other than others—social media sites such as Twitter and

Facebook, for example, are more like each other than they

are to e-newsletters or one-minute speeches—

underscoring the need for theoretical consideration at

the data collection stage of scholarship in this area.

Communicating Across Venues:

Dimensions and Types

Scholars agree that the two parties in Congress are in an

era of deeply polarized relations (Lee 2016;McCarty,

Poole, and Rosenthal 2015). Some studies suggest that the

degree of polarization depends on the setting, with certain

654 Political Research Quarterly 77(3)