Prof. Dr. Beat Hintermann | Léo Picard

Öffentliche Finanzen

Research Topic
In the United States, political campaigns are predominantly negative. They
increasingly involve communications that consist of attacks towards political
opponents, whose policies or characters are framed negatively. This development is
worrisome, as it can undermine socieatal cohesion. Negative campaigns have been
linked to lower trust in politics (Mutz and Reeves, 2005), hateful communication
on social media (Hopp and Vargo, 2017), and affective polarization (Nai and Maier,
2023). In addition to the presence of attacks, the aim and type of the messages
also matter, as personal attacks and uncivil language lead to a lower quality
of the policy debate and stronger affective polarization. The result is a decrease
in electoral accountability and a worsening of the resulting policies that are
implemented in a country. Even though the data for this project is from the USA,
negative campaigning is also an increasingly important element in Swiss politics,
in particular in the context of initiatives (Nai and Sciarini, 2018). Furthermore, in
a recent survey on Swiss citizens, higher levels of affective polarization (which
correlates positively with negative campaigning) have been shown to be associated
with lower satisfaction with democracy (Jansen and Stutzer, 2024).

Description of the Problem
Politicians presumably use negative messages if they believe that this strategy
will help them win an election. However, there is a variation in the use of
negative campaigning, and the determinants of engaging in this practice (and the
“viciousness” of the attacks) are not fully understood. In our project, we aim to
identify the conditions under which we expect more vs. less negative campaigning.
Whereas the previous literature has focused on possible “supply factors” of
negativity (e.g., candidates’ personality or institutional rules), the demand factors
originating from the voters themselves have not been studied. Building on the
political economy literature, we assume that voters’ well-being can be used as
determinants for their voting behavior. Most notably, voters’ personal economic
conditions (which are affected by macroeconomic forces such as recessions and
unemployment) are an important determinant for the vote share of incumbents
(e.g., Bechtel and Hainmueller, 2011). This can be explained by voters attributing
hardship to a culprit who is viewed as being responsible for their loss, and serves as
a disciplining device in the democratic process. In economically difficult times, we
would therefore expect voters to be more susceptible to messaging that someone
(or an entire party and their supporters) is at fault for their situation. In an attempt
to win more votes, politicians might then rely more on negative communications in
these times, compared with more prosperous periods. In our proposed project, we
will empirically examine this theory-derived hypothesis.

Objectives
We propose to use data from a natural experiment to examine if the electorate’s
exposure to a negative economic shock causes politicians to engage in more negative
advertising on TV. Because economic conditions and the demand for negative
campaigning may be codetermined, we use an instrumental variables approach.
More precisely, we will leverage shifts in local economic adversity linked to import
competition from Chinese manufacturing, also known as the “China shock”.
Following its accession to the World Trade Organization in 2001, local exposure
to China’s export surge was linked to higher unemployment (Autor et al., 2014,
Acemoglu et al., 2016) and more income inequality (Autor et al., 2016), leading to
political polarization among the electorate (Autor et al., 2020). Our project would
apply the same identification strategy to study campaign negativity. Measures of the
latter are constructed using the database of political advertising from the Wesleyan
Media Project. This database consists of 90,000 political ads videos aired on TV
from 2000 to 2020, combined with various information (e.g., number of airings and
location, sponsor, political affiliation). Current information on the tone of ads is
limited in scope and definition. To overcome this, we leverage recent developments
in Generative AI technologies, such as OpenAI’s GPT, to transcribe videos’ speech
into text and annotate such transcripts for the presence of attacks. Each attack can
be further analyzed for its aim (i.e., if it targets policies or personal characteristics
of the opponent), and civility (i.e., does it use unhelpful and inflammatory words to
turn the opponent to derision).

Importance, Usefulness and Novelty of the Project
Our project will inform on some determinants of negative campaigning that have
not received much attention in the previous literature, and will thus improve
our understanding of when negative campaigning is to be expected. It will also
implement and evaluate a methodological innovation at the frontier of empirical
research in political economy. For comparable classification tasks, existing studies
already point to excellent performance, comparable to human codings (Gilardi et
al., 2023).