Emotion and reason in political language
By Gloria Gennaro and Elliott Ash
“In politics, when reason and emotion collide, emotion invariably wins.”
In his treatise on Rhetoric, Aristotle suggests that persuasion can be achieved through either logical argument or emotional arousal in the audience. This classic dichotomy between emotions and affect (pathos) on the one side and rationality and cognition (logos) on the other has informed not just philosophy but also the social sciences, and in particular the study of political rhetoric.
In the day-to-day of political communication, politicians constantly decide how to amplify or constrain emotional expression, in service of signaling policy priorities or persuading colleagues and voters. The factors underlying a politician’s choice between emotion and reason are many, and could include individual psychology, political pressures, and institutional incentives. In order to gain a scientific understanding of these factors, a necessary first step is to produce measurements of emotional expression in political communication. In our research presented in a recent working paper, we propose to quantify emotionality in politics using the transcribed text of politicians’ speeches. Our new approach, described in more detail below, uses computational linguistics tools and can be validated against human judgments of emotionality.
Fig 1 – Emotionality in U.S. Congress by Chamber, 1858-2014
Time series of emotionality in the Senate (red) and the House of Representatives (green)
We use a text-based measure to analyze emotionality in the transcripts of 6 million floor speeches from the U.S. Congress over a 156-year period (1858-2014). As shown in Figure 1, emotional expression spikes during times of war. Emotionality has been increasing since the late 1970s, when CSPAN started televising floor debates.
The emotion-reason mix also varies across topics within any given year. Emotion is highest in speeches about patriotism, foreign policy, and social issues, whereas reason is more prevalent in speeches about procedure, government organization, and tax policy. When it comes to tax policy, however, Republicans are more than twice as emotional as Democrats. Republicans appear to use emotion, rather than logic, to defend inequality-enhancing tax policies.
We see that overall, Democrats and Republicans do not differ much in their use of emotional language. However, members of the minority party are systematically more emotional than members of the majority party, a striking trend that we see consistently flip as the party in control flips (Figure 2). During the long term of Democrat control in the second half of the 20th century, Republicans consistently used more emotional language. In turn, after Republicans retook the house in 1994, Democrats became more emotive. Throughout the time series, changes in House majorities correspond to changes in relative emotionality in the two parties.
Comparing between legislators, we see on average more emotive speeches by women, racial minorities, and religious minorities. We also find that on both the left and right ends of the political spectrum, those who tend towards ideological extremes (in terms of voting on policies) are most emotional in their speechmaking (Figure 3). All these relations are roughly constant historically and hold across topics.
Fig 3 - Emotionality and Policy Ideology
The horizontal axis reports the DW Nominate Score, dimension 1; the vertical axis reports the average emotionality score by bin
From these results, we can see some interesting shapes forming in the empirical picture of how U.S. politicians use emotional rhetoric. First, emotion appears in situations of disempowerment, not only when politicians are members of the minority party, but also when they are members of disadvantaged minority groups, whether in terms of gender, ethnicity, and/or religion. Second, politicians respond to conflicts, such as class inequality or ideological polarization, with more emotion. Third and finally, media technology (e.g. television) appears to play a role in amplifying these rousing factors. Each of these findings points to promising avenues for future substantive research in the relation between emotions and politics.
These scientific investigations are now possible thanks to our new approach to measuring emotionality in political language. Our approach builds on recently developed computational linguistics tools, which represent semantic dimensions in language as geometric dimensions in a vector space. The algorithm for this purpose - word embedding - transforms words and phrases into vectors, where similar words tend to co-locate and directions in space (dimensions) correspond to semantically meaningful concepts (e.g. Collobert and Weston 2008).
For our research, we construct a dimension corresponding to reason at one pole and emotion on the other. The starting point is a list of emotion words and a list of cognition words, produced and validated by linguistic psychologists at the University of Texas (Pennebaker et al. 2015). We find these points in word space and then construct the poles as the average vectors for the respective word groups (Figure 4). The relative emotionality of a word is the proximity to the affective pole, relative to the cognitive pole. In turn, the emotionality of a document is the average emotionality across the document’s constituent words. The resulting geometric emotion scale is continuous and doesn't rely on the presence of particular words.
In a human validation where annotators ranked pairs of sentences as more or less emotive, our metric agreed with human judgment over 90% of the time (which is much more often than a word-based measure) (Figure 5). Importantly, the measure provides a valid emotion ranking for the whole history of the Congressional Record back to the 1850s, meaning we can make valid empirical comparisons over long time frames.
From an auxiliary analysis using the Google N-grams dataset, we know that emotionality in broader society has decreased over this period, meaning that the upward trend is specific to politics.
About
Gloria Gennaro is a postdoctoral fellow at the Public Policy Group and Immigration Policy Lab at ETH Zurich. She holds a Ph.D. in Social and Political Sciences from Bocconi University and has held visiting positions at Harvard and NYU. Gloria’s research in comparative political economy explores electoral behaviour in democratic societies using causal inference and computational social science.
Elliott Ash is Assistant Professor of Law, Economics, and Data Science at ETH Zurich. Prior to joining ETH, Elliott was Assistant Professor of Economics at University of Warwick, and before that a Postdoctoral Research Associate at Princeton University’s Center for the Study of Democratic Politics. He received a Ph.D. in economics and J.D. from Columbia University, a B.A. in economics, government, and philosophy from University of Texas at Austin, and an LL.M. in international criminal law from University of Amsterdam. Elliott’s research applies tools from economics and data science to law and politics.