Tailoring the truth – evidence on parliamentarians’ responsiveness and misinformation toleration from a field experiment

Abstract

Populist radical right parties (PRRPs) claim to be particularly responsive to people’s needs and have been identified as a major source of disinformation. The present contribution sets up a field experiment to zoom in on one-to-one communication between voters and their parliamentarians. By drawing on pieces of misinformation that are present among different parties’ supporters, artificial citizen’s requests are sent to all 2503 German federal parliamentarians. In fact, PRRP politicians do not turn out to be more responsive and they are by far more reluctant to reject misinformation. In contrast, parliamentarians of all other parties largely object to misinformation, even if it matches their political positions and is shared by their electorates. In opposition to PRRP politicians who reveal signs of vote-seeking behaviour, established parties’ communication behaviour indicates a high degree of intrinsic motivation.

Introduction

It is the nature of political competition that different parties stress different aspects of the same political issue. Recently, however, political communication research has more specifically turned toward studying the dissemination of false information. The rising interest in this pertinent field of study is first and foremost due to the rise of anti-establishment populist radical right parties (PRRPs) who endorse ‘fake news’ in a twofold way.

First, PRRPs use ‘fake news as a label’ (Egelhofer and Lecheler, 2019: 97) to accuse and discredit certain sources of news. A main characteristic of PRRPs is constituted by a Manichaean discourse that distinguishes between a putatively corrupt elite and the ‘good people’. In this vein, established media and expert knowledge count much less than people’s ‘unfiltered’ experience, ‘gut feelings’, emotions, and common sense (Hameleers, 2022). Since ‘the elite’ putatively ignore the latter and argue based on facts that might be unrelated to citizens’ daily experiences, populists can portray themselves as the people’s only true representation (Mu05e, 2007).

Second, PRRPs have applied ‘fake news as a genre’ by intentionally creating ‘pseudojournalistic disinformation’ (Egelhofer and Lecheler, 2019: 97). Many scholars have convicted populist politicians of strategically spreading false information and polarizing political co01unication (Castanho Silva, 2018; Runciman, 2018). Importantly, a narrow line has been drawn between the deliberate propagation of ‘disinformation’ and the unintentional spread of ‘misinformation’ (Hameleers, 2020: 148). Examples of deliberate disinformation may be the White House press secretary Sean Spicer’s pretension that Donald Trump was received by the largest inauguration crowd in the history of the United States (The Guardian, 2017), Boris Johnson’s famous Brexit campaign claim that the UK sends 350 million Pounds to the EU every week (Full Fact, 2017), or the German Alternative für Deutschland’s (AfD) exaggeration that homicides committed by foreigners in Germany had increased by 685% in 2018 (Tagesschau, 2019). Even these examples, however, reveal the difficulties in differentiating between misinformation and disinformation (Egelhofer and Lecheler, 2019).

Intentionally or not, the rise of radical politicians and their (social) media coverage is often associated with the rise of misinformation (Bennett and Pfetsch, 2018; Berning et al., 2019; Gerstlé and Nai, 2019; Schmidt, 2020). As a matter of fact, PRRP supporters have been strongly exposed to false information, particularly through social media content (Allcott and Gentzkow, 2017; Gunther et al., 2019; Guess et al., 2020). Unsurprisingly, clutching at inaccurate knowledge known as ‘overconfidence’ (van Prooijen and Krouwel, 2019: 159) has been shown to correlate with support for the populist radical right (van Kessel et al., 2021).

In contrast to the well-researched sphere of semi-public social media and the level of misinformation present among PRRP supporters, much less is known about the behaviour of PRRP politicians when being confronted with misinformation in private, direct, or one-to-one communication with their constituents. Recently, such discourses have been analysed in innovative responsiveness experiments based on constituent inquiries: In the US, senators were found to strategically tailor their answers on a controversial issue congruent with the stated constituent’s preferences (Grose et al., 2015). Until today, this strand of literature does little to address the special role that direct voter-politician communication might play for the radical right. The present analysis bridges this gap by carrying out a citizen’s inquiry experiment and asking the following research question:

How do politicians behave when confronted with different pieces of misinformation? Do PRRP politicians stand out in their behaviour in comparison with representatives of other party families?

The present analysis links the PRRP disinformation and misinformation literature that focuses on social media with the experimental approaches that fail to zoom in on the populist radical right behaviour when dealing with misinformation in citizen’s inquiries: A total population experiment of one-to-one citizen-parliamentarian communication of all 2503 national and state parliamentarians is carried out. Misinformation is operationalized for the AfD’s and the Green Party’s core issues – immigration and climate change.

Literature review

Tailoring the truth on the populist right: the German case

For decades, Germany has been interpreted as particularly resilient against the appearance of a significant political competitor on the far right (Bornschier, 2012). The German electoral system follows the principal of proportionality on the national and state level, generally enabling small (and radical) competitors to enter the parliaments. However, a candidate needs to win the relative majority of ‘first votes’ in their constituency to ‘directly’ enter a parliament or parties need to meet an electoral threshold of 5% of the overall ‘second votes’ to nominate a faction of parliamentarians through a pre-defined party list.

In the course of the reorganization of the North-Western European party systems from the traditional redistribution issue to an extremely salient cultural/anti-immigration cleavage (Bornschier, 2010; Hutter and Kriesi, 2019), the German party system finally stretches between a typical populist radical right party and its ‘natural’ opponent – a Green Party. Founded in 2013, within only five years the populist radical right Alternative für Deutschland (AfD) entered the national and all 16 state parliaments. The young party was not only able to draw upon several well-known conservative politicians but also on several experienced staffers that had served for several political competitors (Pinkert et al., 2017). In the federal election in 2021, 16 AfD parliamentarians not only entered the parliament proportionally to the party’s overall share of votes but also received the highest vote share among the candidates in their electoral districts.

Hence, the current German political system represents an interesting object of study to carry out an experiment that focuses on the behaviour of PRRP parliamentarians when faced with misinformation. Within its European populist radical right party family, the AfD represents ‘a case in point’ (Conrad, 2022: 58) that rides the wave of ‘post-truth campaigning’. The party regularly bills itself to be particularly responsive and to represent ‘the true people’ (‘wir sind das Volk’) while being criticized for spreading misinformation by its competitors and the media (Siri and Lewandowsky, 2019). The German PRRP employs an anti-elite rhetoric, promotes a traditional way of life, stages itself as the defender of a homogeneous, nativist nation state, opposes immigration, and denies that climate change resulted from human activities (Schmitt-Beck, 2017; Arzheimer and Berning, 2019). Furthermore, the AfD dominates online communication among German parties and manages to circumvent potentially dividing topics by aggressively focusing on anti-immigration positions (Serrano et al., 2019). Unsurprisingly, ‘disinformation trackers’ reveal that seven of the ten most spread items of disinformation in the German federal election campaign 2017 have been shared by the German PRRP (Sängerlaub et al., 2018). In fact, the AfD is understood to have triggered the rise of disinformation in Germany, which is increasingly perceived as a threat to democracy (Reuter et al., 2019).

On the other cultural end of the political spectrum, the Green Party embraces a multi-cultural society, favours post-materialist values, and stresses Germany’s responsibility in the fight against climate change as their core issue (Bündnis 90/Die Grünen, 2020). In line with their extreme positioning in the German climate change debate, the Greens were accused by the AfD of ‘hysterical climate-crisis screeching’ (own translation, Frindte and Frindte, 2020) as the right-wingers demanded ‘fact-based climate and energy politics’ (own translation, Deutscher Bundestag, 2018).

Hence, the AfD and the Greens promote opposing policy platforms in various topics (see Figure 1) and their electoral success can be clearly attributed to their respective core issues: anti-immigration for the AfD; the fight against climate change for the Greens. Furthermore, AfD politicians regularly claim to be particularly responsive as only they are the people’s true spokesperson. The Green Party also claims a special responsiveness, for example, regarding the grassroots movement Fridays for Futures (own translation, Bundestagsfraktion Bündnis 90/Die Grünen, 2019).

Figure 1. Issue positioning in the German party system.

Source: Chapel Hill Expert Survey (2020).

New experimental evidence: responsiveness tests with parliamentarians

Up to know, little is known about the motivation and cognitive mechanisms behind politicians’ communication behaviour in polarized political systems (Sheffer et al., 2018). In general, legislators are understood to be intrinsically-motivated if their behaviour is driven by their view of how a dutiful politician should behave (Norris, 1997). Intrinsically-motivated parliamentarians are not expected to discriminate between specific groups in the electorate, and selective or strategic communication should be less frequent. In contrast, legislators are understood to be extrinsically-motivated if they are primarily driven by vote-seeking behaviour (Cain et al., 1984). In order to understand the motivation behind representatives’ communication behaviour, it is helpful to disentangle intrinsic and extrinsic motivational drivers (Bénabou and Tirole, 2003). In line with the concept of extrinsic motivation, Butler et al. (2012) compare selective answering to an election campaign’s ‘microtargeting’; Bol et al. (2021) reveal higher response rates if an inquiry includes a personal vote intention or if a candidate is elected ‘directly’ (not through a party list); and Bowler and Farrell (2011) find that legislators elected in smaller districts are more responsive. In fact, extrinsic motivation often seems to dominate the decision of whether a request is answered or not (Vries et al., 2016), crowding out intrinsic motivation (Giger et al., 2020).

Experimental studies that analyse the role of politicians in their direct communication with constituents draft an e-mail or letter with a request or question posed by an artificial citizen to a parliamentarian. This form of ‘contacting’ is a standard way for citizens to address their representatives at low-cost (Teorell et al., 2006; Theocharis and van Deth, 2018). ¹ For Bundestag parliamentarians and their staff who deal with a great variety of different duties, communication with constituents represents one of the core tasks (Bröchler and Elbers, 2001). Response rates to citizen’s inquiries are found to be comparatively high in Germany, ranging between 63 and 79% for Bundestag MPs (Heß et al., 2018; Bol et al., 2021) and 79% for local governments (Grohs et al., 2016). ² It is important to note that parliamentarians in Germany stand on a level playing field in terms of funding. They are free to pick their own staff and design their own communication strategy. The Bundestag offers a monthly budget of 19,913 Euros for each parliamentarian to be spent on their employees, financing the salaries of a total staff of 4,500 people (Deutscher Bundestag, 2015). It remains in the control of individual parliamentarians how many resources are devoted to communication activities. Nevertheless, the literature on the organizational structure of parliamentarians finds that replies in audit experiments differ between politicians and their staff. Also, more professionalization is associated with less discriminatory answers (Landgrave and Weller, 2019).

Costa (2017) summarizes the growing field of experiments (Butler and Broockman, 2011; Broockman, 2013; Grohs et al., 2016; Levine and Glick, 2017): The easier it seems to convince a voter the more responsive elected officials turn out to be – namely if the politician and the voter share relevant characteristics. Unfortunately, these studies remain silent on the treatment of misinformation or disinformation and on the peculiarities of populist radical right politicians. The populist ideology is fundamentally based on the Manichaean division between a corrupt elite and the ‘true people’ and is translated into a specific form of communication: By circumventing established media and promoting direct democracy, populists often claim to take citizens’ concerns more seriously than other politicians (Mudde, 2007; Krämer, 2014; Schürmann, 2022). Hence, the intent of PRRPs to represent voters that feel alienated from mainstream parties and the respective increase in turnout has even been dubbed a ‘corrective for democracy’ (Mudde and Kalwasser, 2017). ³

Given the people-centric populist ideology, the effect of extrinsic motivation on responsiveness should be less pronounced among radical right populists. Since PRRP representatives claim to take people’s concerns seriously, independent of whether a request is based on expert knowledge or empirical evidence (Hameleers, 2020), they should respond less selectively when being ‘contacted’ with flawed pieces of evidence.

H1: In contrast to selective responsiveness to misinformation among established parties, we expect PRRP parliamentarians to be more responsive independent of the stimulus.

Once representatives decide to answer an inquiry, they must deal with the respective content. Again, if responsive democratic representatives followed an intrinsic idealistic conception, all requests would be answered ‘neutrally’ by correcting potentially wrong facts and supplying a political interpretation independent of their party’s issue positioning. If extrinsic motivation mattered, however, parliamentarians might ‘not really care that much about the truth or they care about other things (partisan causes) more than the truth’ (Chambers, 2021: 148).

Hence, the party’s issue positioning could affect the toleration of misinformation. Since a piece of misinformation might imply a certain topic is more pressing or severe than it actually is, it may be tolerated or rejected depending on a party’s political agenda. Finally, a conscious vote-seeking behaviour would be reflected by greater tolerance for disinformation in line with a party’s positioning.

Such an interpretation, however, understates the cognitive constraints that limit politicians’ conscious behaviour. Based on the theory of motivated reasoning (Kunda, 1990), recent research has advanced the psychological understanding of political behaviour and isolated the effect of a confirmation bias (Jakobson, 2010; Knobloch-Westerwick et al., 2015; Westerwick et al., 2017). In a field experiment, Sheffer et al. (2018) reveal that political representatives base their decision-making heavily on framing effects and are more often subjected to choice anomalies than ordinary people. These cognitive biases could drive representatives to unconsciously ignore partisan misinformation in constituent inquiries (Kappes et al., 2020; Körösényi et al., 2022). Both extrinsic vote share maximization (disinformation) and an implicit confirmation bias (misinformation) imply a higher tolerance for flawed facts in line with a party’s (core) issue positioning.

H2: The more closely a misinformation is related to a party’s positioning, the higher the tolerance of respective pieces of misinformation in citizen’s requests.

Since all politicians somehow rely on winning votes, depending on the party’s positioning and the pieces of misinformation, there is a smaller or larger temptation to ignore their constituents’ flawed views on an issue. For parties with an extreme issue positioning – for example single issue pro-climate or anti-immigration parties – it appears even more difficult to resist respective partisan misinformation. What is more, following Mudde’s (2007) understanding of PRRPs, it seems well understandable that respective politicians turn out to be particularly lax when dealing with facts.

First, radical right parties are grounded on a nativist illiberal ideological position. In this sense, the idealistic conception of a politician who truthfully explains pros and cons of an issues to their constituents in order to meet the ideal of deliberative democracy is simply less appealing for PRRP parliamentarians – far beyond their nativist core issue.

Second, the party family is characterized by their populist anti-establishment appeal. Based on Weyland (1999), there is a longstanding tradition in understanding populist representatives as rational actors who employ ‘a series of tactics in order to concentrate power’ (Rueda, 2021: 169). Following the unrestrained ‘people’s will’, translates into less respect for specific moral guardrails. In fact, as populists’ key communication elements, discursive and stylistic simplification can be identified as well as the defiance of expert knowledge (Moffitt, 2018; Hameleers, 2020). By challenging established gatekeepers, they intend to raise concerns about the status quo of the political, societal and economic situation: ‘the communication and mobilization by populist actors build their success on the cultivation of collective grievances’ (Giebler et al., 2021: 914). Deliberately proclaiming a societal crisis is one of the key elements on the populist political agenda (Rooduijn, 2014). Hence, extrinsic motivation would particularly trigger the deliberate toleration of negative disinformation among PRRP politicians.

Taken together, PRRPs’ particular communication behaviour is driven by a radical denial of liberal as well as deliberative democracy and their gatekeepers, in combination with a stark commitment to cherish the unfiltered ‘people’s will’. Furthermore, even if parliamentarians were inclined to correct misinformation, they need to deal with an electorate that might not favour being challenged in their political views. It needs to be acknowledged that extreme political attitudes generally correlate with a ‘black and white’ understanding of complex issues (van Prooijen and Krouwel, 2019). Building upon the findings from social psychology, van Kessel et al. (2021) show that supporters of the populist radical right stand out in their level of misinformation, even compared to non-voters. Challenging prevailing misinformation can be difficult as they ‘engage deeper emotional truths […] that wilfully defy reason’ (Bennett and Livingston, 2018: 135). Given the ‘overconfidence’ (van Prooijen and Krouwel, 2019) that many PRRP supporters exhibit, it is possible that their political representatives face higher electoral costs when correcting negative misjudgements (Chambers, 2021). Radicalism and populism in combination with potentially higher electoral costs leads to the hypothesis that PRRPs are especially inclined to overlook flawed facts beyond their specific issue positioning.

H3: Disinformation tolerance is expected to be particularly pronounced for the populist radical right.

Misinformation operationalization and research design

A major difficulty in operationalizing false information in an experimental setting is the distinction between deliberate disinformation and unintentional misinformation. To avoid confusion in this ‘grey zone’ (Tandoc et al., 2018), the following experiment is limited to quantifiable pieces of misinformation that the different parties’ supporters actually believe in.

Topic-specific misinformation inquiries are designed for the two most controversial issues in the past German elections. Whereas the federal election of 2017 is widely interpreted as a vote on immigration (Korte, 2019) – with the AfD winning a historically high share of 12.7% – the election of the European Parliament in 2019 was dominated by the issue of climate change (Kaeding et al., 2020) – with the German Green Party skyrocketing to a 20.5% election result. Fictional constituent inquiries are drafted around these core issues. The experiment employs the actual misperceptions of party supporters on very general and clearly verifiable facts. To generate comparable results for the political extremes, an immigration vignette employs the numerical overstatement of a negative characteristic of immigrants and a climate change vignette employs the numerical understatement of the progress in fighting climate change (see Figure 2). Dealing with immigration statistics, it needs to be noted that people generally overestimate immigration numbers as well as underestimate their labour market performance and unemployment rates (Alesina et al., 2022). Hence, we need to be cautious when interpreting our results as parties could be comparatively used to flawed facts when immigration is concerned.

Figure 2. Artificial citizen’s request sent out⁴.

Note: Randomized parts are in bold.

As a first stimulus, parliamentarians are asked whether ‘actually’ 48% of immigrants in Germany were unemployed in line with AfD supporters’ perception. The true value was only 14.4% in 2020 (Federal Employment Agency, 2021a). The second stimulus asks whether the current share of renewable energy of total electricity consumption in Germany laid ‘merely’ around 35% in line with the German Green Party supporters’ perception. The true value increased to 45.4% in 2020 (Federal Environment Agency, 2020). Hence, the first misinformation backs the AfD’s anti-immigration positioning in line with the beliefs of their electorate. The second misinformation stimulus provides an argument to accelerate investments in renewable energy in line with the Green Party’s positioning and beliefs of their electorate. Despite the fact that even parliamentarians of established parties might be used to misperceptions of immigrants, they are most likely not used to people expecting every second immigrant to be unemployed.

It needs to be considered that these misinformation treatments differ qualitatively and quantitatively. ⁵ Since the immigration misinformation is quantitatively larger and qualitatively more clearly aligned with the AfD’s positioning than the renewable energy treatment is with the Green Party’s positioning, differences in the party’s communication behaviours over the stimuli need to be interpreted carefully. As a control stimulus, the Germany-wide perceived unemployment rate of 24% is chosen ⁶ – whereas the official average 2020 unemployment rate was 5.9% and increased only slightly over the course of the pandemic (Federal Employment Agency, 2021b).

To increase responsiveness in line with Bol et al. (2021) and Giger et al. (2020), the fictional inquirer pretends to be a constituent, sympathizes with the addressed politician, and offers a potential vote for future elections. To reduce the chances that the request is forwarded to and answered by a central representative from the faction, the inquirer states to live in the parliamentarian’s constituency. The inquiries are designed to be salient but neutral (neither pro nor con regarding the piece of misinformation: ‘is it really true that […]’), to test politicians’ responsiveness and their tolerance for misinformation as generally as possible: The voter pretends to feel insecure regarding the conflicting information that is present in the debate on an issue and asks for the parliamentarian’s assessment. For reasons of comparability, the empirical question refers to the federal level.

To assess the party-level differences of responsiveness and misinformation tolerance in electorate-parliamentarian communication, data of all German national and state parliamentarians is collected including their names, gender, e-mail addresses, the states where the constituencies are located as well as their faction, faction size, government participation by faction, issue positioning and salience by faction as well as upcoming elections in 2021. ⁷ In case such information was unavailable online, the parliament was contacted in order to receive the missing data. Further information on the legislators’ mode of election (directly or through a party list) ⁸ was retrieved from Abgeordnetenwatch. ⁹ To assess whether parliamentarians themselves or their staff answered an inquiry, the name that signed an answer was compared to each inquiry’s recipients.

Finally, a complete mailing list of 703 national and 1800 state parliamentarians of the six major German parties was compiled. ¹⁰ Comparable to Bol et al. (2021), two e-mail addresses were created consisting of common German names and surnames to circumvent potential responsiveness bias in relation to gender or ethnicity (Butler and Broockman, 2011). The common German e-mail service t-online.de served as an e-mail provider.

The e-mails were sent to parliamentarians in two waves to spread the same e-mail content over several days and thus avoid being detected as a fake inquiry or spam by the parliamentarians and their staff. A third of each faction’s parliamentarians received either an immigration, a renewable energy or an unemployment related inquiry. The distribution of issues has been randomized over parties and states. Time spans between the waves were kept to a minimum of one week to reduce the probability of political events that complicate comparability between the waves. To increase responsiveness, the e-mails were sent in January 2021 avoiding parliamentarian holidays and election periods. In sum, three different stimulus e-mails (immigration, renewable energy, unemployment) were sent by two different aliases in two separate waves, yielding a 3 x 2 x 2 experiment set-up. Robustness checks verify that the results do not differ for different alias or waves. ¹¹

The experiment enables an analysis of the two dependent variables of interest: First, the topic-related response rate by political faction (responsiveness); second, the share of parliamentarians who respond to the inquiry but remain quiet on the quantitative question or actively tolerate the piece of misinformation (misinformation toleration). Comparing the level of responsiveness and misinformation with parties’ salience, issue positioning, and parliamentarians’ characteristics, we analyse if political communication follows a strategic rationale and if the PRRPs’ behaviour deviates from established parties.

Ethical considerations

Field experiments that intend to uncover political communication between legislators and voters have recently become fashionable in political sciences (Butler and Broockman, 2011; Broockman, 2013; Grohs et al., 2016; Levine and Glick, 2017; Bol et al., 2021). These analyses are usually based on politicians’ answers to artificial voter inquiries. Hence, in order to simulate real communication between citizens and politicians, the political representatives are deceived and left to believe they face inquiries from actual potential voters. In general, the politicians’ deception in such experiments raises ethical concerns (Butler et al., 2012; Landgrave, 2020; Bischof et al., 2022). The fact that the present experimental setting sends e-mails from fictional constituents in order to reveal politicians’ behaviour when dealing with misinformation is particularly questionable. Any potential critique resulting from the present study of politicians deceiving voters stands on shaky grounds as it is based on deception itself. Unfortunately, the research question is not suitable for a less deceptive audit that includes actual information about the underlying research project as proposed by Landgrave (2020).

When investigating how parliamentarians deal with misinformation, we are particularly interested in the communication behaviour on the political fringes. Therefore, disclosing the requests’ scientific nature is expected to bias the different parties’ response behaviour disproportionally. Deception could also have been reduced by recruiting genuine constituents who voluntarily draft requests (Bischof et al., 2022). Unfortunately, the identification of sufficient citizens who hold representative levels of misinformation and consent to participate in the experiment was impossible. Nevertheless, we believe that the evidence revealed from the outlined experiment outweighs the ethical reservations.

First and foremost, in the context of misinformation, disinformation, and the respective prevailing accusations that different parties face, the experiment sheds light on hidden and important democratic processes. If democracy is endangered (Reuter et al., 2019) or even going to ‘end’ (Runciman, 2018) due to partisanship, misinformation, and finally a degeneration of political culture into self-referential subgroups, there is a pressing need for independent and unbiased research on the issue. In the end, it is in the public interest to understand how politicians tailor the truth to deceive voters. That being said, the findings generated in the experiment will be treated as a public good, made available to the research community and consciously spread through media channels. Perhaps the public discussion on the deliberate toleration of disinformation or the unconscious toleration of misinformation by politicians of different parties has the power to trigger a more responsible way of dealing with citizen’s requests.

Second, of the few ways to analyse direct one-to-one communication, e-mail-based experiments trigger the lowest adverse effects and repercussions. Ethically, it would be preferable to observe communication behaviour in a controlled experimental environment. Unfortunately, such a set-up faces challenges of selection bias, external validity, and the difficulties of recruiting a sufficient number of politicians (Sheffer et al., 2018). In e-mail-based experiments the effort for legislators or their team to answer the stimulus e-mails is low and does not alter their general functional capacities. Legislators receive and answer multiple e-mails each day and have standardized answers for requests in their main political issues. In contrast, a citizen’s request is one of the few chances for constituents to interact one-on-one with their representatives. In case parliamentarians or their staff decide to research the true number demanded in an inquiry, the questions posed require a minimum online search. Those politicians who consider answering too time-consuming face no cost when refusing an answer. In line with Alizade et al. (2021), we respond to any parliamentarian who wants to gather further information by a standardized answer that our request was no longer valid.

Third, we addressed potential privacy protection violations by not asking for any confidential information. We deleted the e-mail accounts immediately after the experimental period and saved the original data on an encrypted hard drive. In order to avoid any inference on individual answers, for publication purposes we aggregate quantitative data to the party level.

Fourth, we answered every response with a standardized expression of our gratitude, to prevent lower responsiveness to following service requests.

Fifth, parliamentarians who participated in the experiment received a debriefing e-mail including the final published article and an invitation to discuss the results.

Lastly, our research outline was approved by the academic ethics commission of the University of Duisburg-Essen and pre-registered at the Open Science Framework.

Results

Responsiveness: ‘defender of the true people’ fail to live up to their own demands

Of the 2503 addressed parliamentarians on the national and state levels, 1283 (51.3%) provided an answer. Independent of how intensely replies engage with the inquirers’ questions, an answer to a citizen’s request was coded as responsiveness and to the contrary, every non-response was coded as a lack of responsiveness. ¹² The response rate of 65.6% for Bundestag parliamentarians lies within the range that has been identified by similar experiments on the national level in Germany (Heß et al., 2018; Bol et al., 2021). The response rate on the state level turns out to be significantly lower (45.7%), raising awareness for structural differences between the different federal layers. Compared to the responsiveness in other mature democracies (Costa, 2017) and despite the specificity of the requests that did not fit into most addressed parliamentarians’ issue specializations, the response rate reveals German parliamentarians’ high willingness to engage with citizens on issue-specific topics.

Figure 3 dives deeper into the responsiveness analysis and provides an overview of response rates by party and citizen’s request. Response rates are the lowest for AfD politicians ¹³ : Only 41.9% of all PRRP parliamentarians provide an answer to the experiment’s citizen’s inquiries. For each of the three different citizen’s requests, Germany’s self-declared ‘defender of the true people’ turns out to be less responsive than members of all other major political parties. Splitting the sample by Bundestag and Länder parliaments does not alter this finding. Regarding one-to-one responsiveness, the AfD seems to be the cause, rather than the saviour, of the putatively missing link between citizens and politicians.

Figure 3. Response rates, by party, in percent.

Source: Own depiction.

Response rates are comparable between the unemployment (54.2%) and the renewable energy (55.9%) query, but significantly lower for the immigration citizen’s requests (44.1%). The responsiveness ordering differs substantially for CDU/CSU, FDP, DIE LINKE and SPD. The Greens come second-to-last in the unemployment and immigration treatments and only manage to leave behind the FDP and left-wing parliamentarians in responsiveness to the renewable energy inquiry. The AfD ranges behind the established parties: Even when being approached by a citizen who is ‘concerned’ about immigration and shares the AfD supporters’ severe overestimation of immigrants’ unemployment rate, only 39.5% of the populist politicians provided an answer ¹⁴ – in contrast to a 44.8% response rate among non-AfD politicians. ¹⁵

Contrary to our expectations (H1), PRRP parliamentarians are not more but less responsive than other politicians. To test for selective response behaviour and control for faction specific differences in the motivational drivers, the following logistic regression models are specified:

$$\Pr {\left( {Responsiveness = 1} \right)_i} = logi{t^{ - 1}}\left( {{\beta _0} + {\beta _i}issue\;salience + {\beta _j}Controls} \right)$$

The binary responsiveness variable represents the dependent variable. We expect varying responsiveness in line with the respective party’s issue salience of our stimuli $i$ (unemployment, renewable energy, or immigration). Party-specific salience of the three issues is extracted from the Chapel Hill Expert Survey (CHES) (2020). ¹⁶ Since national MPs have access to more resources, we control for Bundestag affiliation on the individual level. The same could be possible for government parties and larger factions which we control for on the party-state-level. Signs of extrinsic motivation could be inferred from higher responsiveness in states with upcoming state elections in the coming year or for directly-elected politicians who strongly rely on the support of their constituency. ¹⁷ Since different parliaments potentially supply different levels of resources available to MPs, standard errors are clustered at the state level. Table 1 (Columns 1 to 3) shows the regression results of the three different issue treatments. Hypothesis 1 follows the rationale that parliamentarians are more willing to engage with their party’s core issue. As politicians are regularly approached with questions concerning their core issue, they have almost certainly prepared respective arguments, facts, and policies. Hence, they might be more confident to answer than when asked about less-common issues. Thus, for each regression we expect ${\beta _i} \gt 0$ . ¹⁸ For example, the highest response rate to the renewable energy issue – that is based on a piece of misinformation circulating among Green Party supporters – is expected from the Green Party. ¹⁹ Finally, the subsamples are pooled for a joint regression, including party dummy variables for parliamentarians of the AfD and the Green Party (see Table 1 Column 4):

$${\rm{Pr}}\left( {Responsiveness = 1} \right) = logi{t^{ - 1}}\left( {{\beta _0} + {\beta _1}AfD + {\beta _2}Green\;Party + {\beta _j}Controls} \right)$$

Table 1 Logistic regression models: responsiveness

Robust standard errors in parentheses, clustered at parliament level.

***p < 0.01,

**p < 0.05,

*p < 0.1.

Table 1 reveals no relationship between party-specific issue salience and the response rate. For all three citizen’s requests, there is no statistically significant relationship between issue salience and responsiveness. In contrast, national parliamentarians are more responsive than their state-level counterparts – despite when they are addressed with the immigration misinformation. In general, neither indicators of extrinsic motivation – an election year or having won the constituency ‘directly’ ²⁰ – nor evidence for a higher level of experience – a larger faction or government participation – goes along with higher responsiveness. However, responsiveness is higher in states with an upcoming election if contacted with the renewable energy request. ²¹ Despite the party-specific differences in issue salience, this topic has dominated the German political agenda in 2021 (Forschungsgruppe Wahlen, 2022). Possibly the importance of the climate change issue has triggered higher response rates during the election campaigns.

Finally, Table 1 (Column 4) reveals that ceteris paribus the low responsiveness of AfD parliamentarians is not statistically significant when considering the discussed covariates. AfD politicians are less often national than state parliamentarians, they do not form any government, their factions are smaller than those of other parties, and they win their constituencies less often directly – characteristics that are positively correlated with responsiveness in bivariate regressions. Hence, the AfD’s responsiveness is the lowest, but its differences to established parties are at least partly explained by the party’s peculiar characteristics. Nevertheless, we can reject the populist radical right’s claim of being particularly responsive.

Misinformation toleration: post-truthism on the political right

Of the parliamentarians who answered the citizen’s request, 92% corrected the misinformation and included a numerical number or reference to the official statistics. ²² Most parliamentarians who provide a numerical answer supply the official data or a number that resembles the official statistics closely. However, 103 of the 1283 (8%) parliamentarians’ answers failed to reject the pieces of misinformation. An inquiry is coded to have tolerated a flawed fact if it fails to reject the misinformation. Answers are coded to fail to reject the respective piece of misinformation for different reasons. First, this holds for answers if they simply ignore the empirical question and rather discuss the more generally posed question or an issue unrelated to the inquiry. Second, answers that falsely state there was no official data to evaluate the issue are coded to have tolerated misinformation (‘I can neither confirm nor deny the figures you mention […]. Exact figures are not published’.). Third, answers are coded as a toleration of misinformation if they consent with the inquirer’s proposed misinformation (‘These numbers are frightening and so are the consequences for our society’). The slightest notion of contest or an alternative numerical answer disqualifies a response from being coded as misinformation toleration. This even holds for an AfD politician who states that the unemployment rate ‘cannot be defined correctly right now’ but at the meantime ‘seems implausible’. Hence, this measure of misinformation toleration represents the very conservative lower bound of uncontested misinformation.

Impressively, Figure 4 reveals the party-specific patterns of misinformation tolerance between parties and the stark tolerance of false facts by AfD parliamentarians. Whereas 29.7% of AfD parliamentarians fails to contest the pieces of misinformation presented to them, the share for established parties’ politicians lies around only 5%. With only eight of the 318 SPD respondents tolerating misinformation, social democrats come the closest to the ideal of politicians as a source of unbiased information. Hypothesis 2, which claims that parties rather tolerate misinformation the more closely flawed facts are related to a party’s positioning, has to be rejected. Particularly the Green Party is highly resilient to misinformation even in its core topic. In contrast to underplaying the development of renewable energy in Germany in line with their supporters’ views, 93.5% of the Green parliamentarians contest the respective piece of misinformation. Thus, the AfD’s accusation about the Greens’ ‘hysterical climate-crisis screeching’ (own translation, Frindte and Frindte, 2020) must be rejected. On the other end, four out of ten AfD parliamentarians tolerate or even reinforce the misinformation that the unemployment rate of immigrants ranged around 48% – as stated by their supporters. ²³

Figure 4. Misinformation toleration rate, by party, in percent.

Source: Own depiction.

Finally, the following logistic regressions model the relationship between misinformation toleration and the party’s issue positioning in order to test for strategic behaviour or cognitive biases in politicians’ communication (H2):

$$\Pr {\left( {Toleration\;rate = 1} \right)_i} = logi{t^{ - 1}}\left( {{\beta _0} + {\beta _i}issue\;positioning + {\beta _j}Controls} \right)$$

The individual misinformation toleration variable is employed as the dependent variable. The Chapel Hill Expert Survey (2020) supplies data on issue positioning. Since a special interest in tailoring the truth in line with constituents could evolve from government participation, an upcoming election, or being directly elected, we add respective covariates as controls. Furthermore, we control for faction size and Bundestag affiliation, as these characteristics could indicate a higher level of professionalization and truthfulness. We cluster standard errors on the faction level in each parliament because the local political environment could play a role in dealing with misinformation. Again, we additionally run a logistic regression model for all treatment subsamples, but we also specify a pooled regression model that includes party specific dummy variables to specifically analyse the behaviour of PRRP parliamentarians and their counterparts (H3):

$${\rm{Pr}}\left( {Toleration\;rate = 1} \right) = logi{t^{ - 1}}\left( {{\beta _0} + {\beta _1}AfD + {\beta _2}Green\;Party + {\beta _j}Controls} \right)$$

Table 2 reports the issue-specific (Columns 1 to 3) and pooled (Column 4) regression results. ²⁴ More right-wing parties turn out to have a higher chance of tolerating negative misinformation regarding the labour market, renewable energy, and immigration. The former two findings are surprising and lead us to reject Hypothesis 2. We expected anti-climate change parties to be more ‘flexible’ in dealing with environmental misinformation and left-wing parties to rather tolerate negative labour market misinformation. The regressions show that parties with respective issue positionings tend to shy away from exploiting misinformation in their favour – even if these misconceptions are present among their own supporters. In contrast, the regression results are driven by the AfD’s high misinformation tolerance. Excluding the AfD from the sample, the issue positioning coefficients become statistically insignificant. This also holds for the immigration citizen’s request, where more right-wing (anti-immigration) parties tend to tolerate negative misinformation. ²⁵

Table 2 Logistic regression models: toleration rate

Robust standard errors in parentheses, clustered at parliament-party level.

***p < 0.01,

**p < 0.05,

*p < 0.1.

Table 2 (Column 4) reveals a statistically significant relationship between AfD membership and misinformation toleration: Translated into odds ratios, AfD politicians have a 6.6 times higher chance to tolerate misinformation than parliamentarians of established parties – ceteris paribus. In contrast to Hypothesis 2, misinformation toleration does not stem from a general confirmation bias or motivated reasoning that leads to an unconscious ignorance of partisan misinformation. The insignificance of being directly elected, having upcoming elections, or being better equipped points to intrinsically-motivated politicians who carefully reject misinformation. ²⁶ Hence, the analysis of direct voter-politician communication reveals that for most parliamentarians of established parties, intrinsic motivation to supply correct information dominates the temptation to simply confirm whatever citizens ‘have heard’ – even if these pieces of misinformation would let the respective party’s positioning appear more pressing. In contrast, the extrinsic drivers of misinformation toleration might either fall on fruitful grounds within the populist radical right (disinformation) or be driven by especially pronounced cognitive biases (misinformation) (H3). Finally, the extreme anti-immigration positioning might reinforce but not entirely explain the AfD’s overall misinformation tolerance. The party’s communication behaviour rather needs to be understood following Mudde’s (2007) characterization of the party family’s radical opposition to deliberative and liberal democracy as well as its reluctance to challenge people’s views even if these oppose officially provided statistics.

Conclusion

The prominently-detected dealignment between ‘the elite’ and ‘the people’ (Mair, 2013) puts the spotlight on ‘the action or inaction of political elites and their interaction with the citizens’ (Merkel, 2018). Regarding the (deliberate) spread of misinformation and disinformation, scholars have pointed to the rise of the populist radical right and its anti-establishment ideology (Castanho Silva, 2018; Runciman, 2018; Hameleers, 2020). The present analysis argues that, taken together, the key characteristics of PRRPs shape the party family’s misinformation affinity. First, radicalism drives a fundamental opposition to specific aspects of liberal and deliberative democracy. Second, populism implies an unwillingness to defy the unfiltered ‘people’s will’ as well as an inherent mistrust in established sources of information.

Furthermore, we do not stop at focusing on the populist radical right, but rather ask if communication behaviour follows an inherent topic-specific logic. Parliamentarians’ decisions to answer as well as what to answer might be influenced by different motivational mechanisms. All politicians face time constraints, opportunity cost, choice anomalies (Sheffer et al., 2018) as well as the extrinsic temptation of pretending to understand citizens’ concerns in order to win their votes (Cain et al., 1984).

The experiment was carried out in Germany and reveals the expected division in a proportional voting system that recently rearranged around a salient cultural cleavage. In general, all established parties seem to be rather intrinsically-motivated in their ‘private’ political communication: First, misinformation toleration is low among this group (5.4%) and does not follow a topic-specific pattern. Particularly, the Green Party – even approached with a piece of misinformation on poor renewable energy supply that their supporters believe – entirely lacks signs of motivated reasoning or confirmation bias. Its politicians abstain from tolerating partisan misinformation. Second, responsiveness is high for established parties (52.7%), despite the specificity of the requests that do not match parliamentarians’ individual issue specializations. Higher response rates on the national level, compared to the state level, point to structural differences in the different federal layers of the German political system. Additionally, more than two thirds of citizen’s requests seem to have been answered by parliamentarians themselves and no topic-specific strategic communication behaviour can be detected, not even for parliamentarians who were directly elected in their constituencies.

The opposite is the case for the German populist radical right. Based on their radical and populist ideology, we expected AfD politicians to be more responsive as well as more ‘flexible’ in dealing with misinformation. In line with the ‘affinity’ (Hameleers, 2020: 146) between (right-wing) populism and mis- and disinformation (Waisbord and Amado, 2017), nearly 30% of AfD parliamentarians’ answers fail to contest the pieces of misinformation presented to them. The regression models translate these differences into probabilities and estimate ceteris paribus a 6.6 times higher chance of tolerating misinformation for an AfD politician than for their established counterparts. When being confronted with the overestimation of immigrants’ unemployment rate of 48%, as AfD supporters actually believe, 40% of populist radical right parliamentarians tolerate this piece of misinformation.

Nevertheless, even the extremely high tolerance for immigration-related false facts leaves open whether PRRP communication follows conscious and strategic extrinsic motivation (disinformation) or unconscious motivated reasoning (misinformation). It is up to future research to disentangle the two potential drivers of political misinformation toleration and figure out why the radical right behaves differently. In contrast to the AfD’s people-centric appeal, the party clearly defies its claim to be especially eager to engage with citizen’s inquiries. Although the AfD does not stand out regarding the share of requests answered by staffers, the intuition that the young party’s low responsiveness could be explained by parliamentarians and their staff being rather inexperienced, needs to be carefully monitored during the coming legislative periods.

Furthermore, due to the peculiarities of the German party system, the present experiment has some limitations regarding its external validity. It would be promising to run a comparable experiment in a party system that entails a PRRP that is as established its competitors. Such a set-up could test if the present results are driven by AfD parliamentarians and their staff being less-professionalized. Furthermore, running the similar experiment in a political system that contains a PRRP in government would test whether the present results are shaped by the special oppositional role of the German PRRP. After all, the AfD is the only German party that is neither part of the national nor of any state government.

Finally, it is extremely difficult to identify pieces of misinformation that are comparable across issues. The present study employs stimuli that are qualitatively and quantitatively different. Also, experiments that employ artificial citizen’s requests inherently run the risk of being detected and treated as spam or fraud. External validity could be particularly questioned, if response rates would turn out to be systematically biased between parties.

Despite these limitations, the experiment corroborates the notion present in many European party systems: The threat for the political discourse comes from the populist radical right and the findings are particularly concerning due to the self-referential (analogous) filter bubbles in which PRRP representatives engage with their supporters. After all, people who distrust the political system are less informed, prefer authentic representatives, and do not necessarily favour their views being challenged (Valgarðsson et al., 2021). Typical for a PRRP, their parliamentarians allow ‘totally fabricated scientific data [to] make it past the gatekeepers (Chambers, 2021: 155) and capitalize on ‘society-centred discontent’ (Giebler et al., 2021).

The AfD has been characterized as a ‘case in point’ (Conrad, 2022: 58) of a European populist and radical right party. Accordingly, intrinsic motivation to meet the ideals of deliberative democracy where politicians argue on a truthful level playing field to discover the most valid arguments, might not appeal to parliamentarians (radicalism). Neither do PRRP politicians respect official sources of information that contradict their constituents (populism). Scholarship that monitors increasing discontentment with various political issues and a decreasing ‘factfulness’ of political communication should focus on the radical right end of the political spectrum.