West European Politics

ISSN: 0140-2382 (Print) 1743-9655 (Online) Journal homepage: www.tandfonline.com/journals/fwep20

How do elites view ideology? Analysing

candidates’ associations of left and right

Lukas Warode

To cite this article: Lukas Warode (07 May 2025): How do elites view ideology?
Analysing candidates’ associations of left and right, West European Politics, DOI:
10.1080/01402382.2025.2490886

To link to this article: https://doi.org/10.1080/01402382.2025.2490886

© 2025 The Author(s). Published by Informa

UK Limited, trading as Taylor & Francis
Group

View supplementary material

Published online: 07 May 2025.

Submit your article to this journal

Article views: 1074

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=fwep20
West European Politics
https://doi.org/10.1080/01402382.2025.2490886

Research Note
How do elites view ideology? Analysing
candidates’ associations of left and right
Lukas Warode
University of Mannheim, MZES, Mannheim, Germany

ABSTRACT
Political ideology is a fundamental aspect of politics and a well-researched area
of political science, but difficult to measure. By examining the sentiment of
political elites’ associations with ‘left’ and ‘right’, this study shows the direction
and extent of elites’ attitudes – measured by sentiment analysis – towards ide-
ology by analysing three waves of German Bundestag election candidate sur-
veys (2013, 2017 and 2021). The results show that there is an asymmetry in
attitudes towards political ideology among German candidates. Centre-left and
left-wing candidates consistently view left ideology positively and right ideol-
ogy negatively. Compared to left-leaning candidates, right-leaning candidates,
except the far-right AfD in 2017 and 2021, have less positive attitudes towards
right ideology and less negative attitudes towards left ideology. A key implica-
tion is that the left–right continuum may need to be partially reconsidered or
reconceptualised, as elite attitudes towards ideological poles can significantly
shape political behaviour, communication, and representation.

KEYWORDS Political ideology; left and right; political elites; sentiment analysis; open-ended
survey responses

Political ideology as a system or shared set of political ideas and norms

(Knight 2006) is central to approaching the complexity of political dynam-
ics for both researchers and participants in public discourse. The left-right
dimension is central in the study of political ideology (Downs 1957).
Previous research shows that the terms ‘left’ and ‘right’ have different
understandings across countries (Beattie et al. 2022; Dinas and
Northmore-Ball 2020; Lachat 2018; Tavits and Letki 2009; Zuell and
Scholz 2019), but also within countries (Bauer et al. 2017; Dinas 2012;
Yeung and Quek 2025). The debate about the meaning and identification
of left and right is a polarising issue. Just recently, there was a public

CONTACT Lukas Warode lukas.warode@uni-mannheim.de

Supplemental data for this article can be accessed online at https://doi.org/10.1080/01402382.20
25.2490886.
© 2025 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group
This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://
creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any
medium, provided the original work is properly cited. The terms on which this article has been published allow the
posting of the Accepted Manuscript in a repository by the author(s) or with their consent.
2 L. WARODE

outcry in Germany when the centre-right CDU was described by a public

broadcasting service as a right-wing party.1 Several CDU politicians had
a problem with being placed in the same group – the political right – as
the far-right AfD, while some CDU elites even questioned whether the
CDU was a party of the political right, emphasising the party’s position
in the political centre – the CDU describes itself as ‘die Mitte’ (‘the cen-
tre’). This incident raises several questions that are fundamental to the
usage and measurement of political ideology, since research, e.g. Jolly
et al. (2022), and much of the public would normally agree that the CDU
qualifies as a centre-right party, especially from a historical perspective.
In this research note, I examine how political elites in Germany view
ideology across the political spectrum, by proposing a framework for ana-
lysing the attitudes of German federal election candidates towards left and
right. The case of Germany presents an interesting scenario with a
far-right totalitarian political legacy – Nazism – and a rising far-right
party – the AfD – from which other parties, including the centre-right
CDU/CSU and FDP, try to demarcate themselves. My theoretical expecta-
tion incorporates these contemporary political circumstances and argues
that while political positions matter for attitudes towards left and right,
there is an asymmetry that is skewed towards the left. I expect candidates
from left-leaning parties to view ‘left’ positively and ‘right’ negatively. For
right-leaning, in particular centre-right, candidates, I expect them to view
‘right’ less positively (than left-leaning candidates view ‘left’ positively)
and ‘left’ less negatively (than left-leaning candidates view ‘right’
negatively).
Empirically, I use three waves of GLES candidate surveys for the
most recent federal elections 2013, 2017 and 2021 (GLES 2014, 2018,
2023). I use open-ended response associations with left and right (Bauer
et al. 2017; Jankowski et al. 2023; Zollinger 2024) to measure attitudes
towards ideology and party affiliation, as well as the left–right
self-placement, as indications of political positions. Methodologically, I
apply a validated sentiment analysis framework that incorporates dictio-
nary (Rauh 2018) and transformer models (Guhr et al. 2020; Laurer
et al. 2023).
The results confirm that (a) attitudes towards left and right are system-
atically associated by political position and (b) that the attitudes are
asymmetrically skewed towards the left. The resulting ‘left bias’ is per-
sistent across three waves and different model specifications. Candidates
from left-wing and centre-left parties consistently view ‘left’ as positive
and ‘right’ as negative, while the dynamic is less pronounced for
centre-right candidates, who on average do not view the right as positive
and the left as negative as their left-leaning counterparts. The AfD’s move
 West European Politics 3

to the far right in recent years is reflected in an increasingly positive

(negative) attitude towards the right (left), leading to a clear demarcation
from left and centre-right parties.
This article advances the study of left–right ideology through a novel
focus on elite attitudes. The implications for research and the public are
potentially manifold. Systematic attitude asymmetries could foster elite
polarisation and have potential downstream effects on mass polarisation,
while such asymmetries could also shape public discourse and political
representation, especially if divergent attitudes between representatives
and constituents extend to substantive policy issues.

Elite attitudes towards left–right ideology

Political ideology, often conceptualised as a latent variable, reflects the
idea that political preferences are interconnected within a coherent and
stable system, a perspective that has remained a consistent focus of polit-
ical science research over time (Carmines and D’Amico 2015; Knight
2006). The left–right continuum, central to the study of ideology, func-
tions as a single dimension to structure political competition (Downs
1957) and serves as an overarching ‘superissue’ (Arian and Shamir 1983).
Left and right still serve as relevant foundational markers in political sys-
tems to simplify political dynamics (Cochrane 2015; Jost 2021).
Associations and understandings of ideology, such as ‘isms’ like conser-
vatism, liberalism and socialism (Kurunmäki and Marjanen 2018), carry
certain attitudes towards left and right. While individuals position them-
selves on a left–right scale, they have certain ideas of related isms pro-
jected onto the left–right continuum. Individuals may also have different
associations of what is left and right in terms of isms, e.g. some may see
socialism as the most left-wing manifestation, while for others that place
may be taken by communism (Bauer et al. 2017).
Political elites play a crucial role in this context. First, as elected rep-
resentatives and central figures in public discourse, they play a crucial
role in several aspects of politics, ranging from political representation
and communication to policy-making and governing capabilities
(Körösényi 2018). Second, their use of ideology is characterised by ‘polit-
ical sophistication’, which translates into a more informed and accurate
view of left and right compared to the average citizen (Gallina 2023;
Lupton et al. 2015). However, most research deals with citizens’ (diver-
gent) left–right associations (Bauer et al. 2017; Caprara and Vecchione
2018; Zuell and Scholz 2019; Zollinger 2024), while elites receive less
emphasis (Jankowski et al. 2023), despite their stressed relevance for the
political system. Attitudes towards ideology have not been studied in this
context. This research note addresses this gap by examining elite attitudes
4 L. WARODE

towards left and right, given the continuing relevance of left and right for
political behaviour.

Asymmetrical left–right attitudes: the case of Germany

The German context provides an interesting scenario for studying elite
attitudes towards left-right ideology. As outlined in the introduction, the
German political and party system raises the question of how elites view
ideological labels across the political spectrum. In a default scenario, both
left and right elites should view ‘their’ ideology positively and the ‘oppos-
ing’ ideology negatively – referring to a symmetrical attitude regime – as
shown in Figure 1.
In Germany, attitudes towards the political right are influenced by its
historical legacy (Dinas and Northmore-Ball 2020; Sierp 2014) and efforts
to distance itself from the far-right AfD (Arzheimer and Berning 2019).
Sierp (2014) highlights how the atrocities of the Nazi regime have strongly
stigmatised the label ‘right’, a distinction that is evident in cross-country
comparisons. While direct causal links to Nazism are difficult to establish,
contemporary political culture and discourse, shaped by this legacy, should
influence attitudes towards the right.
While this should give left-leaning elites reasons to be even more neg-
ative towards the right, I also expect centre-right political actors to dis-
tance themselves from the label ‘right’. In terms of party politics, I assume
that the Christian democratic conservative centre-right CDU/CSU and the
liberal centre-right FDP do not want to be grouped together with the
increasingly far-right AfD (see Appendix A in the Online Appendix for
an elaboration). Hence, centre-right elites should have a less positive atti-
tude (0 instead of +) towards right ideology. I expect elites from left par-
ties (centre-left social democratic SPD, centre-left Greens, and left-wing
Linke) to have a stable positive attitude towards the left and a strong
negative attitude towards the right. Therefore, I formulate the asymmetry
hypothesis, which states that right ideology is viewed more negatively
(less positively) than left ideology, as depicted in Figure 2.
The expectation regarding the far-right AfD candidates remains some-
what unclear. There is reason to believe that the party, especially after its

Figure 1. Symmetrical attitudes of left and right elites towards left and right
ideology.
 West European Politics 5

Figure 2. Expected (asymmetrical) attitudes of left and right elites towards left and
right ideology.

move to the far right, should also represent a unique case in the attitude
regime. Therefore, I expect that after 2013, AfD elites will be character-
ised by a distinctly positive attitude towards ‘right’ and a negative attitude
towards ‘left’.
Germany’s political history includes both the far-right Nazi regime and
the more recent, geographically limited, far-left German Democratic
Republic in East Germany. Based on the post-authoritarian ideological
bias mechanism (Dinas and Northmore-Ball 2020) and the right-wing
success in East Germany (Volk 2023), it is crucial to examine whether
East German elites exhibit specific anti-left and pro-right attitudes. The
empirical analysis (Figure A8 in the Online Appendix) includes robust-
ness checks for this mechanism.

Research design and data

This study examines how political elites view left and right, taking into
account their party affiliation and ideological position. Focusing on
Germany, it examines differences in elite attitudes towards left–right ide-
ology based on sentiment analysis of open-ended responses asking respon-
dents how they would describe left and right (see Appendix F in the
Online Appendix for survey item details) from the three most recent
GLES candidate surveys (GLES 2014, 2018, 2023). I compare sentiments
across parties and predict them using left–right self-placement.
Left–right attitude asymmetries can be measured by analysing the sen-
timent of associations across positions on the left–right continuum (see
Figure A1 in the Online Appendix). Symmetry exists when individuals
equidistant from the ideological centre (e.g. 5 and 7 on an 11-point scale
from 1 to 11) have equally positive attitudes towards their own ideology
and equally negative attitudes towards the opposing ideology. When this
balance is absent, as argued earlier, there is an asymmetry or ‘bias’ (see
Appendix B in the Online Appendix for an example-based discussion of
semantic labels across parties).
Sentiment analysis is a valuable tool in political science for detecting
positivity and negativity in political texts. Using validated sentiment dic-
tionaries (Rauh 2018), researchers can comprehensively analyse textual
6 L. WARODE

sources. I adopt an established operationalisation of sentiment that allows

for the detection of sentiment asymmetry (Young and Soroka 2012: 215):

N positive terms − N negative terms

Sentiment Score = (1)
N all terms

The formula for calculating the sentiment score is as follows: the differ-
ence between the number of positive terms and the number of negative
terms is taken, and then the result is normalised by the total number of
terms in the text. This is applicable on the individual (candidate) level,
but also on the party level, where we would aggregate based on either
detected terms (dictionary approach) or predicted labels (classification
model – positive or negative class).
By validating a German sentiment dictionary, Rauh (2018) demon-
strates the applicability of the sentiment score for German political texts.
His ‘augmented approach’, which includes more dictionary words, proves
to be more stable and valid than standard dictionaries. While sentiment
analysis, especially using dictionaries, is often seen as a simple yet effec-
tive method (Rauh 2018; Schwalbach 2022), two key questions emerge:
(1) Can dictionary methods reliably estimate ‘sentiment asymmetry’? and
(2) Is sentiment analysis the appropriate tool for detecting ideological
asymmetries?
For question (1), comparing dictionary results with transformer-based
models and fine-tuned zero-shot classifiers is essential for validation.2
While dictionaries provide interpretability, they lack the complexity of
transformers in handling text beyond bag-of-words. Validation is crucial
in text-based social science research (Grimmer and Stewart 2013). I there-
fore present results from three models: a sentiment dictionary (Rauh
2018), a sentiment transformer trained on German text (Guhr et al. 2020),
and a fine-tuned zero-shot classifier for political text (Laurer et al. 2023).
In order to answer (2), it is important to define the objective of senti-
ment analysis, which is to assess a text’s ‘general polarity on a
positive-negative scale’ (Bestvater and Monroe 2023: 234). A prominent
related but distinct method is stance detection, which aims to indicate
individual positions on concrete propositions (Burnham 2024). The anal-
ysis includes a stance detection model to validate the sentiment analysis
(see Appendix C in the Online Appendix for a longer comparison of
stance detection and sentiment analysis).
The GLES candidate surveys provide a valid representation of leading
political elites, also partially including final elected representatives in par-
liament (see Appendix D in the Online Appendix for details on sample
composition). After removing missing values, the dataset contains 900
respondents in 2013, 700 in 2017, and 735 in 2021.
 West European Politics 7

Results
Figure 3 provides descriptive evidence that generally confirms the asym-
metry hypothesis. The x-axis indicates the sentiment score from negative
(left) to positive (right), while the y-axis presents the parties ordered by
their sentiment score.
Left-leaning parties (SPD, Greens, The Left) consistently show positive
sentiment towards ‘left’ and negative sentiment towards ‘right’ in all three
waves of the survey. Right-leaning parties show mixed results: the CDU
has positive sentiment for the left in 2013, but shifts slightly negative by
2017 and 2021, while maintaining limited positive sentiment for the right.
The FDP’s results are mixed and largely negative for both sides, reflecting
its centrist, liberal ideology, which opposes many typical left- and
right-wing policies. Defining the FDP’s position in this binary left–right
distinction is not easy, as its classical liberal legacy would categorise the
party both as economically right and socio-culturally rather left. The
AfD’s strong positivity towards the right and negativity towards the left in
2017 and 2021 underscores its far-right positioning and is in line with the
radical right normalisation trend (Dinas et al. 2024; Valentim 2024), con-
firming its outlier status in the German party system (see Appendix E in
the Online Appendix for an extensive discussion of the results).
Figure 4 shows the magnitude as the difference between left and right
party sentiment scores (subtracting left from right scores to get an indi-
cation of overall attitudes). The magnitude confirms that candidates from
left parties have a stronger sentiment towards ‘left’ than candidates from
the centre-right CDU and FDP have towards ‘right’. Only the CSU in

Figure 3. Sentiment dictionary (Rauh 2018): left and right sentiment scores by party.
GLES candidate surveys 2013, 2017 and 2021.
8 L. WARODE

Figure 4. Sentiment dictionary (Rauh 2018): magnitude of left and right sentiment
scores by party. GLES candidate surveys 2013, 2017 and 2021.

2013 has a higher magnitude in favour of the right than candidates from
left parties, but this has limited inferential potential as they represent by
far the smallest group of candidates. The magnitude allows us to analyse
the ‘attitude strength’. Since 2017, the AfD is characterised by a particu-
larly positive sentiment towards the right and a negative sentiment towards
the left, which is slightly higher than The Left’s magnitude in favour of
the left in 2017. In 2021, the AfD has the strongest magnitude by an even
larger margin, surpassing The Left on the left side of the ideological
continuum.
In order to validate the dictionary model, Figure 5 presents the party
sentiment scores for the zero-shot model (in the same way as Figure 3),
which classified open-ended survey responses as either ‘positiv’ (‘positive’)
or ‘negativ’ (‘negative’). Figure A6 in the Online Appendix shows the
absolute frequencies of predicted classes by the sentiment zero-shot model
per year and party.
Comparing the two figures, it is clear that there is a strong association
between both the sentiment dictionary application and the zero-shot clas-
sifier. Some patterns are even more pronounced. The centre-right liberal
FDP elites oppose both left and right more strongly than in the sentiment
dictionary model. The AfD’s shift to the right aligns with the sentiment
dictionary model, while the ‘left bloc’ (SPD, Greens and The Left) consis-
tently holds a positive attitude towards the left and a negative attitude
towards the right. The CSU’s results are even slightly higher than those
of the AfD in 2017. While it is to be expected that CSU candidates are
on average more right-wing than CDU candidates, as was seen, for exam-
ple, during the European migrant crisis in 2015, it is important to note
that the CSU forms the smallest group of candidates, which makes the
group more susceptible to inappropriate conclusions.
Figure 6 shows the magnitudes of party sentiment scores, while the
heatmap appears in the Online Appendix (Figure A4). The magnitudes of
 West European Politics 9

Figure 5. Zero-shot classification model (Laurer et al. 2023): left and right sentiment
scores by party. GLES candidate surveys 2013, 2017 and 2021.

Figure 6. Zero-shot classification model (Laurer et al. 2023): magnitude of left and
right sentiment scores by party. GLES candidate surveys 2013, 2017 and 2021.

the zero-shot sentiment model correspond to the magnitudes of the sen-

timent dictionary. There are some minor inconsistencies, for example the
AfD has a stronger ‘right magnitude’ in 2021 than in 2017, while this is
reversed in the dictionary model. The results for the CDU and FDP are
even stronger in comparison to the dictionary model. The CDU always
has a much weaker right-leaning magnitude compared to the strength of
left-leaning magnitudes for parties on the left. While being slightly in
favour of the right in 2013, the FDP has a small left-leaning magnitude
in 2017 and 2021.
Figure 7 compares the correlations between the sentiment dictionary
model, the fine-tuned zero-shot classifier (A) and the German sentiment
classifier (B), restricted to ‘positive’ and ‘negative’ predictions for
10 L. WARODE

Figure 7. (A) Correlations between sentiment dictionary model (Rauh 2018) and
zero-shot classification model (Laurer et al. 2023). (B) Correlations between sentiment
dictionary model (Rauh 2018) and German sentiment model (Guhr et al. 2020). GLES
candidate surveys 2013, 2017 and 2021.

comparability. Strong correlations (R = 0.68 to R = 0.94) validate the

research design, especially the two-class zero-shot model. These results
confirm the suitability of both dictionary-based and transformer models
for analysing ideological attitudes at the party level.3
The analysis so far has focused on the aggregate party level. Figure 8
shows the predicted probabilities of a pooled individual-level logistic
regression across the three waves of the left-right self-placement variable
on positive attitudes towards left ideology (red) and right ideology (blue)
from the zero-shot classification model. The expected relationship holds
for both left and right responses, with left–right self-placement signifi-
cantly predicting sentiment (p < 0.001). Higher right-leaning self-placement
correlates with a lower probability of positive sentiment towards the left
and a higher probability of positive sentiment towards the right.
Figure 9 shows multi-level logistic model predictions for left (top) and
right (bottom) positive sentiment, including party affiliation, survey year
and left-right self-placement. While party differences are not statistically
significant, important trends emerge. The left–right self-placement is in
line with expectations and the pooled logistic model. For the left item,
the intercepts vary by party, with Green candidates showing the steepest
negative slope, suggesting a possible ‘right faction’. Predictions for the
 West European Politics 11

Figure 8. Pooled logistic regression on predicting positive sentiment of left and right.
Predicted probabilities of positive sentiment for left and right open-ended survey
responses based on left-right self-placement. Vertical dotted line indicates mean of
left-right self-placement (4.7). GLES candidate surveys 2013, 2017 and 2021.

Greens with a self-placement above 7 are empirically meaningless, as

Green candidates never have a score higher than 7. For the right item,
AfD and CDU/CSU candidates show stronger positive views of the right
as their self-placement shifts to the right, highlighting the greater impact
of self-placement on the attitudes of right-wing candidates. Moderately
right-wing candidates are less likely to view the right positively than their
more right-wing counterparts.
The Online Appendix also includes a test of the East German anti-left
assumption (Dinas and Northmore-Ball 2020; Volk 2023). There are no
significant results for East German candidates in any model specification
(Figure A8 in the Online Appendix), while the insignificant trend is also
contrary to what was expected. East German candidates are slightly less
likely to have a positive attitude towards the right and slightly more likely
to have a positive attitude towards the left.
Figure A9 in the Online Appendix uses the party sentiment score
framework, but differentiates between candidates who were elected in the
previous federal election and those who were not, and again finds no
difference despite the skewed distribution of the variable (the majority of
candidates were not elected in the last election).
Semantic validation is featured in Figure 10, which shows the most
frequent left and right words associated with predicted sentiment, result-
ing in a 2 × 2 framework. Overall, it reflects that sentiment is associated
12 L. WARODE

Figure 9. Multi-level logistic regression on predicting positive sentiment of left and

right. Random intercepts and random slopes for left–right self-placement, party and
survey year on the candidate level. GLES candidate surveys 2013, 2017 and 2021.

with distinct semantic associations. For example, left-leaning elites associ-

ate ‘Rassismus’ (‘racism’, negative) with the right and ‘Gerechtigkeit’ (‘jus-
tice’, positive) with the left, while right-leaning elites associate ‘Sicherheit’
(‘security’, positive) with the right and ‘Bevormundung’ (‘paternalism’, neg-
ative) with the left. The Online Appendix also features a comprehensive
overview of the most frequent left and right word associations by party
and year (Figure A5).

Discussion
In this research note, I analyse the attitudes of party elites towards left
and right using open-ended survey responses from German Bundestag
candidates (2013, 2017 and 2021). Left elites have a more positive view of
left ideology than right elites have of right ideology. This asymmetry in
attitudes is consistent across all three waves of the survey, with the AfD
since 2017 standing out from this asymmetry with clear positive attitudes
towards the right and negative attitudes towards the left. The results are
stable across four model specifications and are consistent with left-right
self-placement at the individual level.
The study of ideological associations in open-ended survey responses
is becoming increasingly important (Bauer et al. 2017; Gidron and
Tichelbaecker 2025; Jankowski et al. 2023; Zuell and Scholz 2019; Zollinger
2024), focusing primarily on citizens. My study advances this by (1)
examining sentiment-based attitudes and (2) elite (candidate) instead of
 West European Politics 13

Figure 10. Semantic validity of zero-shot sentiment model. Top words per positive/
negative and left/right categories. Top row indicates right semantics, bottom row left
semantics. Left column indicates left-leaning position (negative attitude towards the
right and positive attitude towards the left), right column indicates right-leaning posi-
tions (positive attitude towards the right and negative attitude towards the left). GLES
candidate surveys 2013, 2017 and 2021.

citizen associations across three survey waves. Asymmetrical elite attitudes

towards ideology have implications for political behaviour, communica-
tion, representation and ultimately democratic principles. The attitude
asymmetry towards the left could impact political representation as citi-
zens should expect their representatives to hold corresponding views
towards ideology. While previous research identifies a ‘conservative bias’
among elites, with representatives overestimating the support for right-wing
issue positions of their constituents (Broockman and Skovron 2018), my
study suggests a ‘left bias’ in ideological attitudes. Future studies should
explore vertical association and attitude differences in left–right ideology
between citizens and elites, as perceptions and scale use differ signifi-
cantly (Jessee 2016; Pietryka & MacIntosh 2022; Yeung and Quek 2025).
Elites can strategically emphasise certain ideological attitudes in pub-
lic discourse. Distancing themselves from the label ‘right’ may reflect
genuine disapproval or be strategic communication to attract voters and
14 L. WARODE

gain office. There are potential downstream consequences that can

affect how ideology is viewed and understood by citizens in both sce-
narios. Openly communicated negative attitudes towards right-wing ide-
ology should influence the political behaviour of elites, but also of
citizens, which can be normatively desired or rejected. Analysing atti-
tudes towards left and right, rather than parties, extends the literature
on affective polarisation (Iyengar et al. 2012; Wagner et al. 2023), by
combining ideology and affect (sentiment). While it is expected that
there will be some left–right polarisation of attitudes based on where
individuals fall on the left–right continuum, the question is whether
asymmetry in attitudes actually increases or decreases (affective) polar-
isation compared to symmetrically polarised attitudes (see Figures 1
and 2). Either way, elite left–right attitude polarisation has several
potential consequences, as it can be equated with polarisation of pref-
erences and issue positions, but also reflects different perceptions of left
and right, both of which can exacerbate social tensions and political
animosity.
The bipolar left–right conception of political conflict may need to be
reconsidered in future research, as there is an increasing tripolar trend in
political systems (Zollinger 2024) – left, centre-right and (far-)right
parties – which corresponds to the results of this study. The far-right AfD
increasingly represents a solitary pole alongside the left (SPD, Greens, The
Left) and centre-right (CDU/CSU, FDP, though less consistently), which
is also consistent with the overarching literature on radical right normal-
isation (Dinas et al. 2024; Valentim 2024).
Future research should examine elite left–right attitudes in (1) a mul-
tidimensional way (Carmines and D’Amico 2015; Kitschelt 1994; Kriesi
et al. 2006), investigating whether (dis)like is driven by cultural or eco-
nomic issues – or both – and how these vary across dimensions and
groups; and (2) elite attitudes across countries, taking into account differ-
ences in how left and right are structured across political contexts (Beattie
et al. 2022; Caprara and Vecchione 2018; Dinas and Northmore-Ball 2020;
Tavits and Letki 2009).
Upcoming methodological refinements should focus on studying
semantic meaning and measuring ideological attitudes beyond sentiment,
for example by incorporating emotions (Widmann and Wich 2023) and
advancing stance detection (Bestvater and Monroe 2023; Burnham 2024).
Frameworks using word embeddings offer promising avenues for analys-
ing nuanced political semantics (Rheault and Cochrane 2020; Rodriguez
and Spirling 2022; Rodriguez et al. 2023). The integration of transformers
and large language models (Laurer et al. 2023), with a focus on semantic
context, is crucial for deepening our understanding of ideology.
 West European Politics 15

Notes
1. https://www.deutschlandfunk.de/nach-kritik-an-berichterstattung-ueber-re
chte-parteien-funk-entschuldigt-sich-fuer-fehler-102.html
2. I use the zero-shot model from Laurer et al. (2023) fine-tuned on political
text by Michael Burnham: https://huggingface.co/mlburnham/deberta-v
3-base-polistance-affect-v1.0
3. The Online Appendix provides a complete visualisation (Figure A7) of the
correlations between all models, which further validates the overall agree-
ment across model specifications at the party-year level. The validation
(Figure A7) in the Online Appendix also includes the results of zero-shot
stance detection models (Laurer et al. 2023), which are highly correlated
(R = 0.71 to R = 0.99) with the other model results at the party level. For the
stance detection model, I specified the German labels ‘dafür’ (‘in favour’)
and ‘dagegen’ (‘against’).

Acknowledgements
Previous versions of this article were presented at COMPTEXT 2024, EPSA 2024,
and a graduate school colloquium at the University of Mannheim. I would like
to thank all the discussants and participants for their comments, as well as the
anonymous reviewers of West European Politics and Or Tuttnauer and Marc
Debus for several helpful comments.

Disclosure statement
No potential conflict of interest was reported by the author.

Notes on contributor
Lukas Warode is a PhD candidate in political science at the University of
Mannheim, MZES and member of the Graduate School of Economic and
Social Sciences (GESS). His doctoral research examines how political elites
view the ideological labels ‘left’ and ‘right’ in different political contexts.
[lukas.warode@uni-mannheim.de]

ORCID
Lukas Warode http://orcid.org/0000-0003-0557-0400

References
Arian, Asher, and Michal Shamir (1983). ‘The Primarily Political Functions of the
Left–Right Continuum’, Comparative Politics, 15:2, 139–58.
Arzheimer, Kai, and Carl C. Berning (2019). ‘How the Alternative for Germany
(AfD) and Their Voters Veered to the Radical Right, 2013–2017’, Electoral
Studies, 60, 102040.
16 L. WARODE

Bauer, Paul C., Pablo Barberá, Kathrin Ackermann, and Aaron Venetz (2017). ‘Is
the Left–Right Scale a Valid Measure of Ideology? Individual-Level Variation in
Associations with “Left” and “Right” and Left–Right Self-Placement’, Political
Behavior, 39:3, 553–83.
Beattie, Peter, Rong Chen, and Karim Bettache (2022). ‘When Left Is Right and
Right Is Left: The Psychological Correlates of Political Ideology in China’,
Political Psychology, 43:3, 457–88.
Bestvater, Samuel E., and Burt L. Monroe (2023). ‘Sentiment Is Not Stance:
Target-Aware Opinion Classification for Political Text Analysis’, Political
Analysis, 31:2, 235–56.
Broockman, David E., and Christopher Skovron (2018). ‘Bias in Perceptions of Public
Opinion among Political Elites’, American Political Science Review, 112:3, 542–63.
Burnham, Michael (2024). ‘Stance Detection: A Practical Guide to Classifying
Political Beliefs in Text’, Political Science Research and Methods, 1–18. https://
doi.org/10.1017/psrm.2024.35
Caprara, Gian Vittorio, and Michele Vecchione (2018). ‘On the Left and Right
Ideological Divide: Historical Accounts and Contemporary Perspectives’, Political
Psychology, 39:S1, 49–83.
Carmines, Edward G., and Nicholas J. D’Amico (2015). ‘The New Look in Political
Ideology Research’, Annual Review of Political Science, 18:1, 205–16.
Cochrane, Christopher (2015). Left and Right: The Small World of Political Ideas.
Montreal: McGill-Queen’s University Press.
Dinas, Elias (2012). ‘Left and Right in the Basque Country and Catalonia: The
Meaning of Ideology in a Nationalist Context’, South European Society and
Politics, 17:3, 467–85.
Dinas, Elias, Sergi Martínez, and Vicente Valentim (2024). ‘Social Norm Change,
Political Symbols, and Expression of Stigmatized Preferences’, Journal of Politics,
86:2, 488–506.
Dinas, Elias, and Ksenia Northmore-Ball (2020). ‘The Ideological Shadow of
Authoritarianism’, Comparative Political Studies, 53:12, 1957–91.
Downs, Anthony (1957). ‘An Economic Theory of Political Action in a Democracy’,
Journal of Political Economy, 65:2, 135–50.
Gallina, Marta (2023). ‘Mass Issue Attitudes, Political Sophistication and Ideology:
The European Case’, Acta Politica, 58:4, 856–75.
Gidron, Noam, and Thomas Tichelbaecker (2025). The European Ideological Space
in Voters’ Own Words. Cambridge: Cambridge University Press.
GLES (2014). Candidate Campaign Survey 2013, Survey and Electoral/Structural
Data (GLES). Cologne: GESIS Data Archive. ZA5716 Data file Version 3.0.0.
GLES (2018). Candidate Campaign Survey (GLES 2017). Cologne: GESIS Data
Archive. ZA6814 Data file Version 3.0.0.
GLES (2023). GLES Candidate Study 2021. Cologne: Gesis. ZA7704 Data file
Version 2.0.0.
Grimmer, Justin, and Brandon M. Stewart (2013). ‘Text as Data: The Promise and
Pitfalls of Automatic Content Analysis Methods for Political Texts’, Political
Analysis, 21:3, 267–97.
Guhr, Oliver, Anne-Kathrin Schumann, Frank Bahrmann, and Hans Joachim
Böhme (2020). ‘Training a Broad-Coverage German Sentiment Classification
Model for Dialog Systems’, in Nicoletta Calzolari et al. (eds.), Proceedings of the
Twelfth Language Resources and Evaluation Conference. Marseille, France:
European Language Resources Association, 1627–32.
 West European Politics 17

Iyengar, Shanto, Gaurav Sood, and Yphtach Lelkes (2012). ‘Affect, Not Ideology:
A Social Identity Perspective on Polarization’, Public Opinion Quarterly, 76:3,
405–31.
Jankowski, Michael, Sebastian H. Schneider, and Markus Tepe (2023). ‘How Stable
Are “Left” and “Right”? A Morphological Analysis Using Open-Ended Survey
Responses of Parliamentary Candidates’, Party Politics, 29:1, 26–39.
Jessee, Stephen (2016). ‘(How) Can We Estimate the Ideology of Citizens and
Political Elites on the Same Scale?’, American Journal of Political Science, 60:4,
1108–24.
Jolly, Seth, Ryan Bakker, Liesbet Hooghe, Gary Marks, Jonathan Polk, Jan Rovny,
Marco Steenbergen, and Milada Anna Vachudova (2022). ‘Chapel Hill Expert
Survey Trend File, 1999–2019’, Electoral Studies, 75, 102420.
Jost, John T. (2021). Left and Right: The Psychological Significance of a Political
Distinction. Oxford: Oxford University Press.
Kitschelt, Herbert (1994). The Transformation of European Social Democracy.
Cambridge: Cambridge University Press.
Knight, Kathleen (2006). ‘Transformations of the Concept of Ideology in the
Twentieth Century’, American Political Science Review, 100:4, 619–26.
Körösényi, András (2018). ‘Political Elites and Democracy’, in Heinrich Best and
John Higley (eds.), The Palgrave Handbook of Political Elites. London: Palgrave,
41–52.
Kriesi, Hanspeter, Edgar Grande, Romain Lachat, Martin Dolezal, Simon
Bornschier, and Timotheos Frey (2006). ‘Globalization and the Transformation
of the National Political Space: Six European Countries Compared’, European
Journal of Political Research, 45:6, 921–56.
Kurunmäki, Jussi, and Jani Marjanen (2018). ‘Isms, Ideologies and Setting the
Agenda for Public Debate’, Journal of Political Ideologies, 23:3, 256–82.
Lachat, Romain (2018). ‘Which Way from Left to Right? On the Relation Between
Voters’ Issue Preferences and Left–Right Orientation in West European
Democracies’, International Political Science Review, 39:4, 419–35.
Laurer, Moritz, Wouter van Atteveldt, Andreu Casas, and Kasper Welbers (2023).
‘Building Efficient Universal Classifiers with Natural Language Inference’, arXiv
Preprint, arXiv:2312.17543.
Lupton, Robert N., William M. Myers, and Judd R. Thornton (2015). ‘Political
Sophistication and the Dimensionality of Elite and Mass Attitudes, 1980–2004’,
Journal of Politics, 77:2, 368–80.
Pietryka, Matthew T., and Randall C. MacIntosh (2022). ‘ANES Scales Often Do
Not Measure What You Think They Measure’, Journal of Politics, 84:2, 1074–90.
Rauh, Christian (2018). ‘Validating a Sentiment Dictionary for German Political
Language—A Workbench Note’, Journal of Information Technology & Politics,
15:4, 319–43.
Rheault, Ludovic, and Christopher Cochrane (2020). ‘Word Embeddings for the
Analysis of Ideological Placement in Parliamentary Corpora’, Political Analysis,
28:1, 112–33.
Rodriguez, Pedro L., and Arthur Spirling (2022). ‘Word Embeddings: What
Works, What Doesn’t, and How to Tell the Difference for Applied Research’,
Journal of Politics, 84:1, 101–15.
Rodriguez, Pedro L., Arthur Spirling, and Brendan M. Stewart (2023). ‘Embedding
Regression: Models for Context-Specific Description and Inference’, American
Political Science Review, 117:4, 1255–74.
18 L. WARODE

Schwalbach, Jan (2022). ‘Going in Circles? The Influence of the Electoral Cycle
on the Party Behaviour in Parliament’, European Political Science Review, 14:1,
36–55.
Sierp, Aline (2014). History, Memory, and Trans-European Identity: Unifying
Divisions. New York: Routledge.
Tavits, Margit, and Natalia Letki (2009). ‘When Left Is Right: Party Ideology and
Policy in Post-Communist Europe’, American Political Science Review, 103:4,
555–69.
Valentim, Vicente (2024). The Normalization of the Radical Right: A Norms Theory
of Political Supply and Demand. Oxford: Oxford University Press.
Volk, Sabine (2023). ‘Resisting “Leftist Dictatorship”? Memory Politics and
Collective Action Framing in Populist Far-Right Street Protest’, European Politics
and Society, 24:5, 535–51.
Wagner, Sarah, L. Constantin Wurthmann, and Jan Philipp Thomeczek (2023).
‘Bridging Left and Right? How Sahra Wagenknecht Could Change the German
Party Landscape’, Politische Vierteljahresschrift, 64:3, 621–36.
Widmann, Tobias, and Maximilian Wich (2023). ‘Creating and Comparing
Dictionary, Word Embedding, and Transformer-Based Models to Measure
Discrete Emotions in German Political Text’, Political Analysis, 31:4, 626–41.
Yeung, Eddy S. F., and Kai Quek (2025). ‘Self-Reported Political Ideology’, Political
Science Research and Methods, 13:2, 412–33.
Young, Lori, and Stuart Soroka (2012). ‘Affective News: The Automated Coding
of Sentiment in Political Texts’, Political Communication, 29:2, 205–31.
Zollinger, Delia (2024). ‘Cleavage Identities in Voters’ Own Words: Harnessing
Open-Ended Survey Responses’, American Journal of Political Science, 68:1,
139–59.
Zuell, Cornelia, and Evi Scholz (2019). ‘Construct Equivalence of Left–Right Scale
Placement in a Cross-National Perspective’, International Journal of Sociology,
49:1, 77–95.