Issue ownership in times of crisis: Dynamics of party issue ownership during the COVID-19 pandemic

Abstract

The COVID-19 pandemic became one of the most salient issues for the public in the past few years, so that political parties had to address this issue if they wanted to maximize their vote share and thus the support of the citizens in upcoming elections. Due to the pandemic’s all-encompassing nature, many parties were forced to adjust their policy stances to the multitude of issues facing the public, the economy, and the health care sector. In this contribution, we analyse to what extent parties associated the COVID-19 pandemic with their issue ownership drawing upon German parliamentary parties’ press releases. Using a novel multi-step text analysis approach, we estimate which additional topics parties address in their press releases when publishing their position on the COVID-19 pandemic. We find mixed evidence for our expectation: only in a limited number of prominent issue areas, including domestic affairs, labour policy, macroeconomic issues and civil rights, do the parties we identify as more likely to associate these issues with the pandemic actually make such associations.

Keywords

framing issue ownership press releases language models COVID-19

Introduction

In March 2020, Germany declared an “epidemic situation of national scope” (“epidemische Lage von nationaler Tragweite”)¹, an unprecedented measure that, like in many other countries, had far-reaching societal and political effects. The COVID-19 pandemic rapidly took control of almost all parts of society and quickly monopolized discourse in most countries. While the COVID-19 pandemic was clearly rooted in the topics of global health and health care, the pandemic influenced a wide range of policies related to political, economic, societal and cultural issues. From access to music venues and tourism to civil rights and infrastructure, the term COVID-19 was found in most discussions, even when unrelated to public health. Due to the pandemic’s all-encompassing nature, parties had to focus not only on the pandemic through the frame of health care, but also understand the global pandemic in a more omnipresent way through all the different aspects of life restricted by COVID-19. Despite the pandemic, elections took place in which parties and their candidates competed for the support of the citizens.

This contribution examines parties’ strategies in relating issues with COVID-19. We argue that parties, which we consider as vote-seeking actors that want to maximize the chances to win control over key executive offices (Müller and Strøm, 1999), were keen to associate COVID-19 on topics that fit within their established, long-standing issue ownership. Given that the issue profile of parties matters decisively for the support of parties among the electorate (e.g. Bélanger and Meguid, 2008), addressing policy issues that are of high salience for the respective party’s core voter clientele is likely to maximize the number of votes in upcoming elections.

More precisely, we examine the issue association of COVID-19 by political parties in Germany. In doing so, we focus on the traditional issue profile of German parties and argue that parties should combine statements on COVID-19 with issues that voters associate with the respective parties. Events that come with large-scale media and public attention are associated with an increasing issue salience among the related policy areas. For example, with environmental disasters, McAllister and Bin Oslan (2021) show that these shocks cause an increase in salience of environmental issues, from which in particular ecological parties should electorally benefit, because green parties are often considered competent in dealing with environmental issues. This argument is based on the findings of Bélanger and Meguid (2008) who say that the interaction of salience and issue ownership boosts performance at the polls. However, these examples of exogenous shocks do not fully compare to the influence of COVID-19 on party competition. While with disasters such as bush fires, a clear political issue is being raised, COVID-19 did not focus on one specific issue but became an overarching topic of cross-cutting nature. Therefore, we expect parties to associate COVID-19 with other policy issues, in particular with those that fit with their established issue profile.

Following studies showing that parties tend to associate the pandemic with their preferred issues in election manifestos (Demler, 2022), we evaluate our theoretical framework on the basis of a novel dataset covering all press releases of German parties between 2020 and 2022. We utilise a thorough multi-step text analysis process (Figure 1) that integrates several stages and models, including (a) an unsupervised dividing of larger documents into manageable units, (b) a keyword-based pre-filtering, (c) training and applying sentence transformers to discern COVID-related from non-relevant content and (d) supervised topic classification.

Figure 1.

Our universal pipeline for analysing the issue ownership dynamics in political communication texts, such as press releases, through a multi-step approach that involves creating management units, filtering out relevant segments related to the crisis of interest (e.g., COVID-19), and conducting issue-based issue association analyses.

To evaluate our argument, we employ a clear identification strategy that allows us to isolate the patterns of issue association during the COVID-19 pandemic. Specifically, we leverage variation in parties’ historical issue profiles and the timing of pandemic-related communications to assess how parties selectively link COVID-19 to issues they are traditionally associated with. By combining comprehensive press release data with a multi-step text analysis pipeline, our approach ensures that observed associations are not driven by general increases in communication activity, but rather reflect strategic issue linking. This explicit identification strategy enables a precise assessment of the interaction between issue ownership and crisis-driven political messaging.

Following this new approach, the findings indicate mixed support for our hypothesis. German parties were more likely to associate the pandemic with issues they are traditionally associated with in only a few but prominent cases like domestic affairs, labour policy, macroeconomic issues or civil rights issues. To derive this finding, we present the case and the theoretical framework in the next sections. Thereafter, we introduce the data and estimation techniques. The fourth section presents the findings in a descriptive and multivariate manner. The final section concludes by summarising the findings and outlining ideas for future research.

The COVID-19 pandemic

Agenda setting is especially challenging in times of crises (Wenzelburger et al., 2019) and can force parties to shift towards adaptive strategies (Conti et al., 2018). These challenging shocks can have many forms, such as environmental circumstances. For example, natural disasters have often been described as having an impact on policy setting regarding environmental salience (Skidmore and Toya, 2002), but also human made catastrophes, such as the Fukushima nuclear accident, showed that while some parties and governments did a 180° turn on policy preferences, other parties did not respond at all to the exogenous shock (Meyer and Schoen, 2017; Müller and Thurner, 2017). By examining these occurrences, research can develop a deeper understanding of how political parties respond to unexpected events, which may involve altering their policy stances or emphasising certain issues in response to external shocks and their effects (Traber et al., 2018).

In this contribution, we focus on the issue association with the COVID-19 pandemic by political parties in Germany. More specifically, we evaluate, first, to which degree parties associate their public statements on COVID-19 with other policy issues and, secondly, if these associations follow specific patterns that are rooted in the longstanding policy profile of parties. During the pandemic, the German Federal Government and state governments took measures to control the spread of COVID-19, which can be divided into three phases: pre-COVID-19, spread and containment, and contagion mitigation (Bobba and Hubé, 2021). The pre-COVID-19 phase in Germany lasted until the outbreak was officially detected in China in December 2019. The second phase saw a relatively rapid transition to regulatory intervention to contain COVID-19, and there was a consensus that returnees from risk areas had to undergo a 2-week quarantine, and social distancing became a general rule of behaviour in Germany. The third phase, resulting in a relaxation of restrictive measures, started on the 6^th of May 2020, and common standards for infection control were agreed upon. Government action was also characterised by several structural measures aimed at protecting the healthcare system from possible overburdening by COVID-19 patients. The BMG (Federal Ministry of Health/Bundesministerium für Gesundheit) had already started the extensive procurement of personal protective equipment (PPE), especially masks, by the end of the first phase. The percentage of free intensive care beds was never less than 34% during the time from the 16^th of April to the 30^th of June 2020, and individual German states have made their resources available for the reception of COVID-19 patients from France and Italy. Hence, the COVID-19 situation in Germany saw a lockdown that resulted in a slump in the economy and affected people significantly in almost all dimensions of daily life (Bobba and Hubé, 2021).

Owning and associating issues

The concept of issue ownership has its origins in the seminal works of Petrocik (1984, 1996). Walgrave et al. (2012) have further developed this idea, dividing issue ownership into two parts: “associative” and “competence”. According to Walgrave et al. “associative issue ownership refers to a party being linked to an issue in people’s minds” (2012: p. 773), and is particularly relevant when studying political parties that focus on a single issue. This clearly distinguishes the concept of issue ownership from the necessity of competence (especially proven expertise through government experience) on an issue. The ownership of an issue by a political party stems from its ideological preferences, which may come from historical or institutional backgrounds. Thus, according to Walgrave and De Swert (2007), issue ownership goes back to the cleavage theory by Lipset and Rokkan (1967) and follows Klingemann et al. (1994: p. 24) thesis that ”parties sustain an identity that is anchored in the cleavages and issues that gave rise to their birth”. On a similar note, Petrocik (1996) argues that issue ownership is a stationary existence of the party, related to class or another cleavage that defines their electorate. However, this can be severely influenced by competence, for example through poor government performance, or by not emphasising an issue for a long-term period (Petrocik, 1996). Yet, according to Petrocik (1996), some issues are “performance issues” and, following Walgrave and De Swert (2007), some issues can be “on lease” and are thus more volatile. From past literature, we know, however, that usually parties are reluctant to change their policy positions significantly (Harmel and Janda, 1994; Laver, 2005). This is also true with issue ownership (Seeberg, 2017).

Political parties cannot only focus on issues that are relevant to their own ideology, as, for example, through election campaigns other issues might become salient. Yet, parties may choose to emphasize some issues while ignoring others. Previous research has shown that parties selectively (de)emphasize some issues to increase their electoral support (e.g., Budge and Farlie 1983; Petrocik 1996). However, in many election campaigns, we find parties confronted with an exogenous shock like an environmental tragedy, an economic crisis, international conflicts, or a pandemic, and therefore the parties have to respond to the new electoral context (Budge and Farlie, 1983; Sigelman and Buell, 2004). In these situations, parties can make use of issue framing (Vreese, 2005) or issue reframing (Lefevere et al., 2019). Issue framing refers to strategically emphasizing an issues’ subdimension (Hänggli and Kriesi, 2010). A step further is taken when it comes to issue reframing. In this case, parties, when discussing an issue, avoid focusing on the smaller details within the specific policy area and instead present the issue in relation to other policy domains. Using this method, parties can refocus on issues that help them to win an election or gain votes. As they are likely to be incentivised to use issue reframing in situations where their primary issue ownership is not salient, we argue that it is likely that they will use their reframing of the salient issue closer to their own issue ownership. While we cannot directly test whether parties reframe issues, we examine whether their associations with owned issues increase when they simultaneously address the main issue on the agenda, that is, in our case, the COVID-19 pandemic. Although this does not capture reframing in a strict sense, it allows us to assess whether parties shift attention toward topics that voters typically associate with them, thereby reflecting their strategic behaviour.

Talking about COVID-19 and other policy issues

We are interested in understanding if and to what degree political parties connected and reframed the COVID-19 pandemic and its implications in a way that COVID-19 fits into their respective issue profile. We know that parties try to adopt policy positions that are closely related to voter preferences (see e.g., Ezrow et al., 2011). Furthermore, parties are willing to emphasize policy issues during election campaigns that are salient for voters in general and for the likely supporters of a party in particular (see, e.g., Klüver and Sagarzazu 2016; Spoon and Klüver 2014). The reasoning of adopting such a strategy is that it should help to maximise the voters’ support for the respective parties in upcoming elections. Consequently, parties should stick to their long-standing policy profile when developing positions on newly emerged issues such as COVID-19 and its implications for other policy domains. If a party would completely change its policy profile because of an external shock like the COVID-19 pandemic and would not relate its programmatic views to its established positions on other important policy issues, then voters perceiving this behaviour might attach a lower degree of trustworthiness to the respective party, which should decrease the chances that the party maximizes its vote share in the next election (Greene and Haber, 2015).

A key element of parties’ electoral campaigns and legislative behaviour in-between elections, for instance by initiating bill proposals or by giving speeches in parliament, is highlighting issues that they ‘own’ and where they are perceived as competent by the voters. Such policy domains could, for instance, cover long-term issues that voters associate with specific parties and where a significant share of voters considers the respective parties as competent to solve problems related to the issue domain (e.g., Petrocik, 1996). Referring to theories of interest-based voting, we expect that parties and their representatives combine those policy issues with effects and implications of the COVID-19 pandemic that are of key interest for their likely voters (Baumann et al., 2021). This strategy should both increase turnout among this group and the chances that likely voters cast a ballot for the respective party.

We should therefore see that parties - when speaking about the pandemic - should tend to speak also from their own issue ownership perspective and thus combine their statements on COVID-19 with policy areas that they own and in which they are considered to be competent by their likely voters.

Which political party should combine which issues when responding to the challenges and implications of the COVID-19 pandemic? We use the literature on portfolio allocation as a starting point to develop some specific expectations. We know from several studies that parties try to win control over those cabinet posts which deal with policy issues that are of key interest for their likely voters. Once controlling a cabinet post that is salient for a party and its supporters, the respective party has better chances to getting its policy views implemented, which should increase its chances in maximizing its vote share in upcoming elections. Browne and Feste (1975: p. 533) argued that “some portfolios are viewed by parties as being especially important for the purpose of reinforcing the loyalty of certain extra-parliamentary clientele groups on which they depend for the maintenance of their parliamentary standing”. Relying on studies of party families in Western Europe and historical studies of groups that traditionally support certain parties and cleavages, as well as analyses of supporters’ attitudes and of programmatic statements, Budge and Keman (2003: p. 95) distilled a ranking of portfolio types for five central party families (Conservatives, Liberals, Religious, Socialists, Agrarians; see also Bäck and Debus 2016; Bäck et al., 2011). Applying the ’party family approach’, one could argue that representatives from specific party families combine those policy issues with the COVID-19 pandemic that are of higher importance (or ‘saliency’) for their likely voters than other parties. This should help in mobilizing the parties’ likely voters for upcoming elections. Moreover, if parties are historically “linked” to or “associated” with a policy domain, they have incentives to integrate the respective issue when reacting to a newly emerged issue like COVID-19 because party leaders would run the risk of being accused of ignoring the traditional policy goals of a party or even of “betraying party ideals” (Budge, 2015: p. 770).

Based on these considerations, we expect that parties relate the COVID-19 pandemic in the public statements to those issues that they ‘own’, that is, issues that their likely voters associate with the respective party and where these parties are considered particularly competent by the voters. Following the ideological background of parties, which is reflected in their preferred cabinet posts, we can derive more specific expectations (see also Krauss et al., 2025; Seeberg 2017). More precisely, we expect:

• (Former) communist, socialist or social democratic parties to combine COVID-19 policies more often with labour and welfare policy, which would allow parties of this ideological family to highlight their profile on social security issues including housing and health care,

• Conservative and Christian democratic parties to combine the effects of the pandemic with issues related to domestic affairs and interior policy, so that these parties can demonstrate their preference for a law and order policy,

• Liberal parties to focus on financial, economic and citizen rights issues when discussing COVID-19 because these policy domains allow liberal parties to emphasise their preferences for free markets and individual liberties,

• Green parties to concentrate on environmental, energy and climate issues when speaking about COVID-19 as these topics form the ”backbone” of the policy profile of ecological parties and

• Parties of the far right to stress law and order issues as well as immigration when presenting their positions on COVID-19 because these issues are at the core of the ideological profile of far right-wing parties.

According to several studies, parties from these ideological families do not only try to win control over the related cabinet posts in coalition negotiations, but also strongly emphasise these issues in their election manifestos and campaign communication in European democracies (e.g., Bäck et al., 2011; Däubler et al., 2024). In addition, voters tend to relate these issues to the respective parties and attach a higher degree of problem-solving capacity on the respective issues to the parties. Highlighting these issues in their daily communication during an exceptional time period like a pandemic could be a strategy to frame the issues related to the pandemic towards the ’classic’ profile of a party, which should strengthen its chances to convince voters in upcoming election campaigns.

Transferring these theoretical considerations and general expectations to the German case requires a brief discussion on the recent patterns of German party competition. Franzmann et al. (2020) argue that present-day politics of Germany is characterised by competition among parties on socio-cultural issues, leading to centrifugal party competition. However, there is still an ongoing ideological conflict on socio-economic matters, where the socialist Left and the Social Democratic Party (SPD) focus on traditional redistribution and welfare concerns, whereas the Christian Democrats (CDU/CSU) and in particular the liberal Free Democrats (FDP) emphasise, by contrast, the advantages of a free-market ideology. Additionally, the Alternative for Germany (AfD) and Alliance 90/The Greens mainly concentrate on socio-cultural issues. Analyses of election manifestos on the federal and state level additionally indicate that the CDU/CSU put special emphasis on law and order issues, while the far-right AfD concentrates also on immigration issues (Bräuninger et al., 2020: p, 57–59) in their election manifestos. These considerations imply that the socialist Left and the SPD should connect their COVID-19 policies mostly with labour and welfare policy, while the Christian Democrats should tend to combine their statements on the pandemic with domestic affairs. The fiscally conservative and socially liberal Free Democrats should highlight economic policy and citizen rights issues when talking about COVID-19, while the Greens, who are strongly associated with a green policy agenda (see e.g. Debus and Tosun, 2021), should aim at combining issues related to COVID-19 with environmental issues. Finally, we should see a combination of COVID-19 issues with immigration issues in case of the far-right AfD (Lehmann and Zehnter, 2024).

Data and methods

The data we use are press releases from the German political parties

, and

² that have been released during the time of the COVID-19 pandemic (2020–2022). We obtained the press releases from the parties’ federal press office websites (see the appendix for more details on the crawling), with the exception of the Greens who provided us with the press releases via the Heinrich Böll Stiftung. Table 1 shows the number of press releases in our dataset for each party.

Table 1.

Number of press releases before (B) and after (A) filtering.

							∑
B	1635	1969	967	281	2270	1278	8400
A	268	503	139	221	326	204	1661
B	19.46%	23.44%	11.51%	3.35%	27.02%	15.21%	100%
A	16.13%	30.28%	8.37%	13.30%	19.63%	12.28%	100%

Multi-step approach

Our data processing pipeline functions as follows: (1) We break down each press release into smaller segments, ensuring that each one maintains coherence concerning its topic. This segmentation is essential since we cannot presume that every press release covers a single topic. Analysing segments of semantically related sentences rather than entire documents is expected to enhance the accuracy of subsequent processing steps. (2) After segmenting the data, we aim to identify all press releases discussing the COVID-19 pandemic. To ensure comprehensive coverage, we first apply a dictionary-based filtering method using existing keywords from Thamlitz (2022), Donovan (2021), Beck et al. (2021) and Rieger and Nordheim (2021) with a filtering of similar words (see Table 2 for keyword statistics)³: We compiled a final list of 326 keywords related to the pandemic to form our candidate dataset. (3) We enhance precision and train a classifier that differentiates between press releases that only mention the pandemic and those that delve into one or more pandemic-related aspects in detail. Consequently, we generate a dataset of press releases from each party concerning the COVID-19 pandemic, which serves as the foundation for our analysis.

Table 2.

Number (#) of keywords sourced from various origins to initiate the keyword-filtering process, displaying a wide array of generic and specific keywords utilized to generate an initial collection of relevant segments from the press releases.

Dictionary	#num	Examples	%
Thamlitz (2022)	279	Maske, Impfstoff, …	84.29%
Rieger and Nordheim (2021)	8	Virus, SARSCoV2, …	2.42%
Donovan (2021)	40	Lockdown, Shutdown, …	12.08%
Beck et al. (2021)	4	Corona, Pandemie, …	1.21%

(4) We identify the topics addressed in the press releases. As discussed in the theoretical section of this paper, we expect that political parties will mainly focus on the pandemic in relation to the policy issues they hold responsibility for. To test our hypothesis, we need to collect information regarding the topics for each segment in the press release dataset. For this, we employ a machine learning classifier specifically designed to predict policy issues within the framework of the Comparative Agendas Project (CAP)⁴. The data has undergone manual annotation and includes a set of policy issue labels such as Health, Education, Civil Rights, Immigration, Law and Crime, among others. By enhancing our press release corpus with these policy issues, we will be able to analyse how various parties frame the pandemic. We will now explain each step of our methodology in further detail.

Step 1: Press release segmentation

To segment our press releases into semantically coherent segments, we apply the unsupervised segmentation algorithm of Glavaš et al. (2016). The method computes a semantic relatedness score between sentences, based on static word embeddings, and then constructs a semantic relatedness graph of the document where the individual sentences are nodes in the graph and the edges between the nodes are weighted, based on the similarity score. By applying this algorithm, the press releases are automatically divided into semantically related segments⁵.

Step 2: Keyword-based filtering

The primary objective of the keyword-based (pre-)filtering stage is to exclude all press releases unrelated to the COVID-19 pandemic. Our keyword dictionary, which is informed by several existing dictionaries, aims to achieve high recall (Beck et al., 2021; Donovan, 2021; Rieger and Nordheim, 2021; Thamlitz, 2022). The dictionary not only includes terms that refer to the pandemic itself but also words that are associated with the issues, i.e., words that often co-occur in documents about the pandemic. Our dictionary includes a total of 326 keywords. Next, we gather all segments of press releases featuring a minimum of two keywords from the dictionary. Following the filtering process, our dataset contains 1661 press releases. Table 1 presents the statistics for the dataset after keyword filtering.

The output of the keyword-based filtering step is a corpus of press releases that might be on the topic of the COVID-19 pandemic. Please be reminded that the keyword-based method was tuned for recall, but not for precision. We now apply a second filtering step and train a machine learning classifier to predict which of the press release segments that have been extracted in step 1 are actually about COVID-19-related topics.

One might ask whether the keyword-based filtering was necessary, given that we now apply a second filtering step. The reason why we had to first filter out the majority of irrelevant press releases is two-fold. First, when creating a training and validation set for the classifier, we randomly sampled press releases from our dataset. If we would have skipped the first filtering step, then the majority of the press releases would have been on topics unrelated to our research focus, meaning that in order to obtain a sufficiently large number of training instances, we would need to annotate a much larger subset of the data. The second reason for our approach is that supervised machine learning methods are well known to pick up spurious correlations in the data. Therefore, it is crucial that the training data does include positive and negative instances that contain keywords that might be related to COVID-19. Otherwise, the classifier would be prone to lexical bias, simply learning to focus on a small number of words like ‘pandemic’ and ‘COVID-19’ instead of learning the more fine-grained nuances that distinguish relevant from irrelevant press releases.

To train and validate our classifier, we first created a smaller test set with 100 segments from our press releases corpus that we used to measure inter-coder reliability. Each of these segments has been annotated by three trained coders,⁶ following the instructions in a codebook.⁷ We obtained a high inter-rater agreement with 90.7% (average percentage agreement among the annotators) and a Cohen’s Kappa of 80.6 (average of the pairwise agreement among the annotators, see Table 3)⁸. As the agreement between the coders was sufficiently high, we then annotated an additional set of 1417 segments of press releases, where each segment was annotated by a single coder. The final dataset includes 790 positive (i.e., relevant) samples and 727 negative instances (press release is not relevant for COVID-19).

Table 3.

Agreement analysis among annotators on relevant and irrelevant segments, assessed through Cohen’s Kappa and percentage agreement, demonstrating a high level of consistency across annotation pairs (a1-a2, a2-a3, a1-a3).

Annotator pair	Percentage overlap	Cohen’s Kappa
(Annotator 1, annotator 2)	91%	0.81
(Annotator 3, annotator 2)	91%	0.79
(Annotator 1, annotator 3)	90%	0.81
avg	90.7%	0.80

Step 3: Classify relevance

After the annotation had been completed, we split the 1417 samples into a training and development set. For testing, we use the 100 samples that have been annotated by all three coders. We then trained a neural machine learning model on the data. As our annotated sample is relatively small, with about 1500 instances, we used SetFit (Tunstall et al., 2022) with the ‘deepset/gbert-base’⁹ model as the main monolingual model for German. SetFit is a few-shot learning method designed to work well with limited labeled data, which is important given our small training set. It has shown good results in various text classification tasks even with only a few dozen examples per category, which led us to choose it for our experiments. SetFit, as shown in Appendix Figure 10, builds on pretrained Bidirectional Encoder Representations from Transformers (Devlin et al., 2019) and uses a two-step training process based on contrastive learning. First, the model learns to tell similar and different texts apart by seeing pairs of texts and identifying whether they belong to the same or different categories (positive pairs vs. negative pairs). This helps the model create useful numerical representations of texts, making similar texts closer and dissimilar ones farther apart in a high-dimensional space. In the second step, these representations are used as features to train a simple classifier head for the task (see Appendix for more details).

Our trained model achieved 90% accuracy and an F1-macro of 89.7% on our test set (n = 100), with a precision of 89.2% and a recall of 90.8%. For the positive class of relevant press releases, F1-micro is 91.3 (precision: 96.3%, recall: 86.8%). We applied this model to all press releases in our data, in order to identify relevant documents for the issue-based framing analysis. This resulted in a dataset with 4228 segments of press releases by the six parties, all addressing the COVID-19 pandemic. The distribution of relevant vs. irrelevant segments for each party can be seen in Table 4.

Table 4.

Total count of press releases (#) for each of the five parties (CDU/CSU, FDP, AfD, SPD and Left), highlighting the distinction between relevant (related to corona) and irrelevant (no reference to corona) content as classified by our trained domain-adapted sentence transformer classifier, along with the difference (Δ) computed as (Irrelevant − Relevant).

COVID
Party	Relevant	Irrelevant	Δ	#num
	717	2351	1634	3068
	1122	3464	2342	4586
	415	1868	1453	2283
	445	2489	2044	2934
	1023	4493	3470	5516
	506	1734	1228	2240
∑	4228	16,399	12,171	20,627

Step 4: Predicting policy issues

To investigate which issues different parties associate with the COVID-19 pandemic, we need to augment our data with information on the policy issues discussed in the press releases. For this, we use the CAP topic classifier developed by Klamm et al. (2022). The classifier is also based on Bidirectional Encoder Representations from Transformers (Devlin et al., 2019) and has been adapted to the political domain by applying continuous pretraining (Tunstall et al., 2022), i.e., the pretrained transformer model has been further trained on debates from the Bundestag, using a Masked Language Modelling objective. After the domain adaptation step, the model has been fine-tuned for the task of topic prediction in the framework of the Comparative Agendas Project, using the dataset of Breunig et al. (2023).

In a cross-domain evaluation, Klamm et al. (2022) demonstrated the robustness of the classifier by transferring it from annotations within the Comparative Agendas Project to Parliamentary Debates from the German Bundestag. The overall performance in a completely new domain was 70.2%, showing promising results for using this classifier in a new domain setting. They found that the performance dropped around 7% for out-of-domain prediction compared to in-domain evaluation, where the best performing model achieved a 76.8% performance. It is a well known problem in NLP that applying classifiers across domains and text types often leads to performance loss (Daume Iii and Marcu, 2006; Gururangan et al., 2020). Additionally, the appendix contains a further broader investigation of the classifier’s performance in our setting for press releases.¹⁰

Issues distribution

Figure 2 shows the distribution of topics in press releases related to COVID-19. We can see a difference in the distribution of topics in the two samples. The topic most frequently emphasized in the COVID-19-relevant press releases is Health. The topic most frequently used in the press releases not addressing the pandemic was International followed by Government.

Figure 2.

Distribution of CAP topics across relevant and irrelevant COVID-19 segments identified in the previous step, illustrating the dominance of each topic within our dataset.

Figure 3 shows that the Greens, as expected, associated COVID-19 with Environment issues, while the AfD, also in line with our expectations, linked its COVID-19 press releases predominantly to Immigration. The Left demonstrates a strong connection to Labour and Defence, and, together with the SPD, to Social issues. Domestic matters are primarily associated with the CDU and SPD. In contrast to the other parties, the FDP does not show a strong association with any particular issue. With the exception of the FDP, the German parties tend to associate COVID-19 with those topics that they 'own' and where they are often considered particularly competent by the voters.

Figure 3.

Overview of the highlighted issues (e.g., defence) concerning the COVID-19-related segments in the press releases from the , , , , , and . The figure shows the distribution of the expressed topics by party. For example, in 5.44% of all topics addressed by the Linke, the topic is defence.

Analysis

In our analysis, we investigate if parties are more likely to associate the pandemic with issues that they own rather than issues that they do not own. In so doing, we estimate simple logit models on the basis of all press-releases that focus on COVID-19. Each of these press releases form the unit of observations in the following analysis. The dependent variable provides information on whether a press release combines the pandemic with one of the discussed other policy areas. Our independent variables are dummy variables that identify the parties of interest, that is, CDU/CSU, SPD, FDP, AfD, Greens and The Left. According to our theoretical expectations, we expect that a party should be more likely to combine their press releases on Covid-19 with a policy area that fits with its respective traditional profile. We control for the years in which the press release was published. The results of the logit models are presented in detail in the appendix.

For the , we find no significant effects of them emphasising the environment significantly more than their competitors during the pandemic. This shows that the Green Party did not refer to its main policy domain - environmental issues - when presenting their statements on COVID-19 to the public by means of their press releases. There is no statistically significant difference to the other parties according to the results of the logit model presented in Figure 4 below and in the appendix.

Figure 4.

Effects of party-specific authorship on mentioning environmental policy in press releases on COVID-19.

As we argued above, the radical right AfD has issue ownership on immigration policy. The AfD emphasizes these issues in normal times, more than other parties. Therefore, during COVID-19, the AfD should use their issue ownership on immigration to associate it with the COVID-19 pandemic in order to benefit electorally through more exposure to a topic where they are perceived to have more competence than their competitors. Figure 5 shows that there is no empirical evidence for this expectation. We can see that the AfD does not emphasize immigration more significantly more than the other parties that were represented in the German parliament during the COVID-19 pandemic.

Figure 5.

Effects of party-specific authorship on mentioning immigration issues in press releases on COVID-19.

We identified domestic issues as the policy domain that own. We are evaluating if the CDU/CSU mentions this issue, which covers interior policies like law and order issues, more in the context of the pandemic than other parties. We find indeed that the CDU/CSU emphasised domestic issues more in the context of the COVID-19 pandemic than the Greens and the socialist Left party (Figure 6). The results also indicate that the Christian Democrats associate Covid-19 more with domestic policy issues than AfD and FDP, albeit the effect barely fails standard levels of statistical significance. Given that CDU/CSU and SPD formed the federal government until December 2021 and thus most of the time when the pandemic structured daily life, it is not surprising that also the Social Democrats associated domestic issues with the pandemic in their press releases.

Figure 6.

Effects of party-specific authorship on mentioning domestic policy in press releases on COVID-19.

In Figure 7, we turn to the issues that and own, that is, labour and welfare policy. We find that the AfD talked significantly less about labour and welfare policy during the pandemic than the SPD and the Left. The Left party actually focuses more on labour policy than the SPD. There is no significant difference between the emphasis on labour policy in press releases on COVID-19 between SPD, CDU/CSU, FDP and Greens. A similar pattern emerges when concentrating on welfare policy issues. It is only the AfD which puts less attention on welfare issues in its press releases on COVID-19 than the SPD. There are no statistically significant differences between the remaining parties in the amount of welfare issues they integrated into their press releases that concentrated on the pandemic.

Figure 7.

Effects of party-specific authorship on mentioning social and labour policy in press releases on COVID-19.

In case of the , we find evidence for our expectation that the German liberals associated citizens rights more than most other parties with COVID-19 (Figure 8). In comparison to the FDP, the CDU/CSU, The Left and the Greens emphasized civil liberties significantly less, whereas there is no difference between FDP, SPD and AfD in associating civil liberties with the pandemic in their press releases on COVID-19. Also on macroeconomic policy, an issue that the fiscally conservative FDP traditionally stands for, we find that the liberal party associates economic issues more with the pandemic than CDU/CSU and the Left, but the FDP is not statistically significant from the SPD, Greens and the AfD in their degree to combine economic issues with COVID-19 in their press releases.

Figure 8.

Effects of party-specific authorship on mentioning civil rights and macroeconomic policy issues in press releases on COVID-19.

Overall, we can see that parties in some cases tend to associate issues they are traditionally linked with the COVID-19 pandemic in their press releases. However, we should note that this pattern is far from being robust. There is no clear dominance of one party in the policy area it traditionally stands for when analysing the content of the press releases all major German parties made during the pandemic. Instead, a party that traditionally stands for one policy issue tends to highlight this issue in its press releases on the pandemic together with at least one other party. This was, for instance, the case in civil rights issues which not only the FDP, but also SPD and AfD emphasised more than the other parties, or in the case of domestic policy which SPD and AfD associate to a similar degree as the CDU/CSU with the pandemic and its implications.

Conclusion

In this contribution, we examined the relationship between parties redefining issues during the COVID-19 pandemic and their own issue ownership. As the pandemic was a cross-cutting issue that spread into most areas of life, the parties had incentives to talk about issues not directly related to public health while still emphasising the effects of COVID-19 on other areas. We argued that parties should be more likely to focus on issues that are closer to their own ideology and, thus, their longstanding issue profile. Our theoretical argument was based on multiple reasons. As the COVID-19 pandemic was an all-defining exogenous crisis, voters are likely to need reminding of what parties stand for. Therefore, parties should associate the pandemic with issues that voters typically connect with them and rely on as informational shortcuts. Browne and Feste (1975: p. 533) found evidence for a similar pattern with regard to the allocation of cabinet posts: “some portfolios are viewed by parties as being especially important for the purpose of reinforcing the loyalty of certain extra-parliamentary clientele groups on which they depend for the maintenance of their parliamentary standing”.

We found, however, only mixed evidence to suggest that, during the COVID-19 pandemic in Germany, parties were more likely to associate the pandemic with issues from their own ideology. Only in a few but prominent cases like domestic affairs, labour policy, macroeconomic policy or civil rights issues, the parties we identified as being more likely to associate the respective issues with the pandemic actually did so, mostly together with at least one other party. Future research should investigate to what extent the actual patterns of portfolio allocation in the government may contribute to the identified patterns in parties’ press releases, as varying ministerial responsibilities may contribute to the quantity of press releases independently of the respective parties’ issue profile.

There are further incentives for future research on how parties associate issues with policy domains that they traditionally stand for or “own”. For instance, government parties are more likely to be punished by voters for poor economic performance than opposition parties. Being in office also poses a risk to the party’s image of competency, especially during times of crisis (Stokes, 2019). One could argue that governing parties and opposition parties exhibit distinct issue-salience dynamics during the COVID-19 pandemic. While opposition parties should maintain a consistent focus on their core issues, governing parties may need to shift their priorities when addressing the crisis.

In addition, this contribution illustrated how a multi-step text analysis approach can be effectively employed in this context, leveraging recent advancements. Furthermore, we introduced a new dataset containing press releases that covers a time period not represented in current datasets, alongside a segmentation tool for these releases. We also provide open-source trained models for identifying relevant segments within press releases, as well as a domain-adapted model specifically for this domain. Together, these resources should serve as a foundation for others to explore the relationship between how parties reframe issues over time and how they connect newly emerged issues with their own ownership.

Our contribution extends beyond party responses to COVID-19 and can be applied to a wide range of crises, such as the diverse responses to the Ukraine crisis (Masch et al., 2023). This broader applicability allows future research to examine potential differences in issue strategies and variations between opposition and government parties, which will become more prominent due to omnipresent polycrises.

Supplemental material

Supplemental material - Issue ownership in times of crisis: Dynamics of party issue ownership during the COVID-19 pandemic

Supplemental Material for Issue ownership in times of crisis: Dynamics of party issue ownership during the COVID-19 pandemic by Christopher Klamm, Marc Debus, Simone Ponzetto, Ines Rehbein, Sarah Wagner in Party Politics.

Footnotes

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

Marc Debus disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Support for this research was provided by the German Research Foundation (Project number: 256967640) and by the Mannheim Centre for European Social Research (MZES). Simone Ponzetto and Ines Rehbein acknowledge funding from the German Research Foundation (DFG) as part of the UNCOVER project (PO1900/7-1, RE3536/3-1).

ORCID iDs

Christopher Klamm

Marc Debus

Simone Paolo Ponzetto

Sarah Wagner

Supplemental material

Supplemental material for this article is available online.

Notes

Author biographies

Christopher Klamm is an interdisciplinary PhD student at the University of Mannheim and a research associate at the Cologne Center of Comparative Politics, working at the intersection of Natural Language Processing and Computational Political Science.

Marc Debus is Professor of Political Science, Comparative Government, at the University of Mannheim, Germany. His research concentrates on intra-party politics, party competition, governing in multi-party cabinets, and legislative decision-making.

Simone Ponzetto is Professor of Information Systems at the University of Mannheim, Germany. He is a member of the Data and Web Science Group, where he leads the Natural Language Processing and Information Retrieval group. His research interests lie in the areas of knowledge acquisition, text understanding, and the application of natural language processing methods for research in the digital humanities and computational social sciences.

Ines Rehbein is a postdoctoral researcher at the University of Mannheim, Germany. Her work is focussed on the application of Natural Language Processing methods to research questions in the Computational Humanities and Computational Social Sciences.

Sarah Wagner is a Lecturer in Quantitative Political Science at Queen’s University Belfast. Her research focuses on political parties and their strategic approaches, with a particular emphasis on radical left parties and their voter dynamics.

References

Bäck

Debus

(2016) Political Parties, Parliaments and Legislative Speechmaking. Palgrave Macmillan.

Bäck

Debus

Dumont

(2011) Who gets what in coalition governments? Predictors of portfolio allocation in parliamentary democracies: who gets what in coalition governments? European Journal of Political Research 50(4): 441–478. https://doi.org/10.1111/j.1475-6765.2010.01980.x

Baumann

Debus

Gross

(2021) Strategic issue emphasis in parties election campaign statements. Party Politics 27(3): 515–527. https://doi.org/10.1177/1354068819864091

Beck

Lee

J-U

Viehmann

, et al. (2021) Investigating label suggestions for opinion mining in German Covid-19 social media. In: Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers). Association for Computational Linguistics, pp. 1–13.

Bélanger

Meguid

(2008) Issue salience, issue ownership, and issue-based vote choice. Electoral Studies 27(3): 477–491. https://doi.org/10.1016/j.electstud.2008.01.001

Bobba

Hubé

(2021) Populism and the Politicization of the COVID-19 Crisis in Europe. Springer International Publishing.

Bräuninger

Debus

Müller

, et al. (2020) Parteienwettbewerb in Den Deutschen Bundesländern. Springer.

Breunig

Guinaudeau

Schnatterer

(2023) Policy agendas in Germany A database and descriptive insights. Journal of Legislative Studies 29(4): 485–497. https://doi.org/10.1080/13572334.2021.2010395

Browne

Feste

(1975) Qualitative dimensions of coalition payoffs: evidence from european party governments, 1945-1970. American Behavioral Scientist 18(4): 530–556. https://doi.org/10.1177/000276427501800406

10.

Budge

(2015) Issue emphases, saliency theory and issue ownership: a historical and conceptual analysis. West European Politics 38(4): 761–777. https://doi.org/10.1080/01402382.2015.1039374

11.

Budge

Farlie

(1983) Explaining and Predicting Elections: Issue Effects and Party Strategies in Twenty-Three Democracies. Taylor & Francis.

12.

Budge

Keman

(2003) Parties and democracy: coalition formation and government functioning in twenty states. In: Comparative European Politics. Oxford Univ. Press.

13.

Conti

Hutter

Nanou

(2018) Party competition and political representation in crisis: an introductory note. Party Politics 24(1): 3–9. https://doi.org/10.1177/1354068817740758

14.

Däubler

Debus

Ecker

(2024) Party campaign statements and portfolio allocation in coalition governments. West European Politics 47(1): 216–227. https://doi.org/10.1080/01402382.2022.2140397

15.

Daume Iii

Marcu

(2006) Domain adaptation for statistical classifiers. Journal of Artificial Intelligence Research 26: 101–126. https://doi.org/10.1613/jair.1872

16.

Debus

Tosun

(2021) The manifestation of the green agenda: a comparative analysis of parliamentary debates. Environmental Politics 30(6): 918–937. https://doi.org/10.1080/09644016.2020.1864130

17.

Demler

(2022) Im Schatten von Covid-19: Thematisierungsstrategien der Parteien in der Bundestagswahl 2021. Zeitschrift für Parlamentsfragen 53(2): 287–303. https://doi.org/10.5771/0340-1758-2022-2-287

18.

Devlin

Chang

M-W

Lee

, et al. (2019) BERT: pre-training of deep bidirectional transformers for language understanding. In: Proceedings of the 2019 Conference of the North. Association for Computational Linguistics, pp. 4171–4186.

19.

Donovan

(2021) The political exploitation of Covid-19: the AfD as challenger party and the impact on parliament. Zeitschrift für Parlamentsfragen 52(4): 824–843. https://doi.org/10.5771/0340-1758-2021-4-824

20.

Ezrow

De Vries

Steenbergen

, et al. (2011) Mean voter representation and partisan constituency representation: do parties respond to the mean voter position or to their supporters? Party Politics 17(3): 275–301. https://doi.org/10.1177/1354068810372100

21.

Franzmann

Giebler

Poguntke

(2020) It’s no longer the economy, stupid! issue yield at the 2017 German federal election. West European Politics 43(3): 610–638. https://doi.org/10.1080/01402382.2019.1655963

22.

Glavaš

Nanni

Ponzetto

(2016) Unsupervised text segmentation using semantic relatedness graphs. In: Proceedings of the Fifth Joint Conference on Lexical and Computational Semantics. Association for Computational Linguistics, pp. 125–130.

23.

Greene

Haber

(2015) The consequences of appearing divided: an analysis of party evaluations and vote choice. Electoral Studies 37: 15–27. https://doi.org/10.1016/j.electstud.2014.11.002

24.

Gururangan

Marasović

Swayamdipta

, et al. (2020) Don’t stop pretraining: adapt language models to domains and tasks. In: Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, Online. Association for Computational Linguistics, pp. 8342–8360.

25.

Hänggli

Kriesi

(2010) Political framing strategies and their impact on media framing in a Swiss direct-democratic campaign. Political Communication 27(2): 141–157. https://doi.org/10.1080/10584600903501484

26.

Harmel

Janda

(1994) An integrated theory of party goals and party change. Journal of Theoretical Politics 6(3): 259–287. https://doi.org/10.1177/0951692894006003001

27.

Klamm

Rehbein

Ponzetto

(2022) FrameASt: a framework for second-level agenda setting in parliamentary debates through the lense of comparative agenda topics. In: Proceedings of the Workshop Parlaclarin III Within the 13th Language Resources and Evaluation Conference. European Language Resources Association.

28.

Klingemann

H-D

Hofferbert

Budge

(1994) Parties, policies, and democracy. In: Theoretical Lenses on Public Policy. Westview Press.

29.

Klüver

Sagarzazu

(2016) Setting the agenda or responding to voters? Political parties, voters and issue attention. West European Politics 39(2): 380–398. https://doi.org/10.1080/01402382.2015.1101295

30.

Krauss

Praprotnik

Gross

(2025) Issue ownership and government participation: explaining parties’ issue attention at the subnational level. Party Politics 31(6): 941–951. https://doi.org/10.1177/13540688241306732

31.

Laver

(2005) Policy and the dynamics of political competition. American Political Science Review 99(2): 263–281. https://doi.org/10.1017/s0003055405051646

32.

Lefevere

Sevenans

Walgrave

, et al. (2019) Issue reframing by parties: the effect of issue salience and ownership. Party Politics 25(4): 507–519. https://doi.org/10.1177/1354068817736755

33.

Lehmann

Zehnter

(2024) The self-proclaimed defender of freedom: the AfD and the pandemic. Government and Opposition 59(4): 1109–1127. https://doi.org/10.1017/gov.2022.5

34.

Lipset

Rokkan

(1967) Party Systems and Voter Alignments. Free Press.

35.

Masch

Demel

Schieferdecker

, et al. (2023) Shift in public opinion formations on defense, energy, and migration: the case of Russia’s War against Ukraine. International Journal of Public Opinion Research 35(4): edad038. https://doi.org/10.1093/ijpor/edad038

36.

McAllister

Bin Oslan

(2021) Issue ownership and salience shocks: the electoral impact of Australian bushfires. Electoral Studies 74: 102389. https://doi.org/10.1016/j.electstud.2021.102389

37.

McHugh

(2012) Interrater reliability: the kappa statistic. Biochemia Medica 22(3): 276–282.

38.

Meyer

Schoen

(2017) Avoiding vote loss by changing policy positions: the Fukushima disaster, party responses, and the German electorate. Party Politics 23(4): 424–436. https://doi.org/10.1177/1354068815602145

39.

Müller

Strøm

(1999) Conclusions: party behavior and representative democracy. In: Müller

Strøm

(eds) Policy, Office, or Votes? 1 ed. Cambridge University Press, pp. 279–310.

40.

Müller

Thurner

(2017) The Politics of Nuclear Energy in Western Europe. Oxford University Press.

41.

Petrocik

(1984) Explaining and predicting elections: issue effects and party strategies in twenty-three democracies, by Ian Budge and Dennis J. Farlie. Political Science Quarterly 99(1): 105–106. https://doi.org/10.2307/2150269

42.

Petrocik

(1996) Issue ownership in presidential elections, with a 1980 case study. American Journal of Political Science 40(3): 825. https://doi.org/10.2307/2111797

43.

Rieger

Nordheim

(2021) Corona100d. Docma Working Paper;4. Publisher: Technische Universität Dortmund.

44.

Seeberg

(2017) How stable is political PartiesÄô issue ownership? A cross-time, cross-national analysis. Political Studies 65(2): 475–492. https://doi.org/10.1177/0032321716650224

45.

Sigelman

Buell

(2004) Avoidance or engagement? Issue convergence in U.S. presidential campaigns, 1960–2000. American Journal of Political Science 48(4): 650–661. https://doi.org/10.2307/1519925

46.

Skidmore

Toya

(2002) Do natural disasters prevent long-run growth? Economic Inquiry 40(4): 664–687. https://doi.org/10.1093/ei/40.4.664

47.

Spoon

J-J

Klüver

(2014) Do parties respond? How electoral context influences party responsiveness. Electoral Studies 35: 48–60. https://doi.org/10.1016/j.electstud.2014.04.014

48.

Stokes

(2019) The permanent refugee crisis in the Federal Republic of Germany, 1949. Central European History 52(1): 19–44. https://doi.org/10.1017/s0008938919000025

49.

Thamlitz

(2022) Erstellen Eines Deutschen Twitter-Datensatzes Über COVID-19. Master Thesis. University of Mannheim.

50.

Traber

Giger

Häusermann

(2018) How economic crises affect political representation: declining party voter congruence in times of constrained government. West European Politics 41(5): 1100–1124. https://doi.org/10.1080/01402382.2017.1378984

51.

Tunstall

Reimers

UES

, et al. (2022) Efficient few-shot learning without prompts. https://arxiv.org/abs/2209.11055

52.

Vreese

(2005) News framing: theory and typology. Information Design Journal 13(1): 51–62. https://doi.org/10.1075/idjdd.13.1.06vre

53.

Walgrave

De Swert

(2007) Where does issue ownership come from? From the party or from the media? issue-Party identifications in Belgium, 1991-2005. Harvard International Journal of Press/Politics 12(1): 37–67. https://doi.org/10.1177/1081180x06297572

54.

Walgrave

Lefevere

Tresch

(2012) The associative dimension of issue ownership. Public Opinion Quarterly 76(4): 771–782. https://doi.org/10.1093/poq/nfs023

55.

Wenzelburger

König

Wolf

(2019) Policy theories in hard times? Assessing the explanatory power of policy theories in the context of crisis. Public Organization Review 19(1): 97–118. https://doi.org/10.1007/s11115-017-0387-1

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

1.06 MB