Who tweets, and how freely? Evidence from an elite survey among German politicians

Twitter has become one of the primary platforms for politicians to interact with the public. Consequently, research into politicians’ Twitter usage has proliferated with attempts at measuring increasingly complex concepts such as ideology or policy attitudes. So far, many of these studies either implicitly or explicitly assume that politicians’ Twitter accounts are operated by politicians themselves and that politicians are free to present their “true” attitudes and positions. We conducted an elite survey in Germany and present evidence that these assumptions only partially hold true. In our sample, only around a third of Twitter accounts are operated by the corresponding politician alone. In our view, this is a conservative estimate and should further decrease as political elites’ social media strategies professionalize over the coming years. We also find that most politicians state that there are no party guidelines regarding Twitter and that their tweets are not checked by a central authority in the party. We discuss the implications of our findings for research on social media in general.


Introduction
Ever since the introduction of Twitter in 2006 and its subsequent adoption by politicians, political elites' usage of the platform has become an increasingly popular topic in political science. Twitter data is often regarded as a valuable resource to measure individual behavior, preferences, or positions of politicians. This literature, either implicitly or explicitly, commonly makes two assumptions crucial to most of the corresponding studies. First, scholars implicitly assume that political elites control their accounts themselves (e.g., Barberá and Zeitzoff, 2018;Fazekas et al., 2021;Stier et al., 2018). Even if studies sometimes acknowledge that the staff of Members of Parliament (MPs) often co-curate and sometimes even exclusively control their social media communication (Kelm et al., 2019;Kreiss et al., 2018), it is usually only mentioned in passing. In a survey among German MPs administered in 2012, 59% of responding MPs answered that they controlled their social media accounts exclusively or predominantly themselves, and 91% of them said they felt they had full or predominant control over their accounts (Meckel et al., 2013). Given the growth of Twitter, it is time to provide more recent evidence.
Second, it is often assumed that politicians can express their "true" positions and preferences on Twitter as parties don't act as gatekeepers in this arena (Castanho Silva and Proksch, 2021a;Ceron, 2017;Sältzer, 2022 whether politicians use Twitter only by themselves or if they delegate this task to their staff and whether parties control the Twitter activity of MPs to any degree. We present the results of an original survey fielded among members of the German Bundestag in 2020 to explore these assumptions. We targeted the 709 members of the 19th German Bundestag, all of whom were invited to take part in the survey via email. 1 The number of questionnaires collected (and thus completely filled out) is 183, which corresponds to a response rate of 25.8%. 2 While this number might seem low, it is comparable to other, similar elite surveys (e.g., Kelm et al., 2019;Meckel et al., 2013).
Looking at the participation rate of the individual parties (see Supplementary Appendix, Table A4), the Free Democratic Party (FDP) has the highest response rate with 42.5%, followed by the parties The Left (37.7%) and Alliance 90/The Greens (34.3%). 23.2% of all members of the alliance between the Christian Democratic Union and the Christian Social Union (CDU/CSU) and 21.1% of all members of the Social Democratic Party of Germany (SPD) took part in the survey. The Alternative for Germany (AfD) had the lowest response rate, with 12.4%. Of course, such response rates are a challenge with regard to representativeness, that is, the question of whether they allow conclusions to be drawn about the target population. Supplementary Table A5 in the appendix shows a detailed comparison of sample and population characteristics (e.g., age, gender, seniority, and electoral district type). When conducting the survey, we assumed that employees would often complete the questionnaire. Hence, we explicitly asked about this. As expected, a large proportion of the questionnaires were answered by staff (around 77%), which is expected from elite surveys.

Who is operating the account?
Survey participants were asked who actually controls the MP's account: the MP herself, staff, or both the MP and staff. Figure 1, Plot 1 presents the results: in most cases, both the MPs and their staff use the account (56%). In 33% of the cases surveyed, only the MPs use the account, and in 11% of the cases, it is only their staff. The insight for us is that many (re)tweets from MPs' accounts do not originate from MPs personally. We also assume that staff members that manage a Twitter account are often responsible for other social media accounts. This empirical insight is central to research on social media since the account is often implicitly equated with the politician's person. A natural follow-up question concerns the number of people that have access to the Twitter account and use it. As seen in Figure 1, Plot 2, a significant proportion of Twitter accounts are operated by two or more people. Higher numbers of account operators may also reflect a higher degree of professionalization, and a significant share of accounts in our sample, 37%, is operated by three or more people.

Are there guidelines on Twitter?
Furthermore, it is of great interest whether there is coordination within parties regarding social media usage. Such coordination could, for example, include guidelines issued to the members of parties or a central instance that coordinates and controls politicians' social media posts. As shown in Figure 2, across the different parties, the majority of respondents in our sample state that there are no rules or guidelines for dealing with Twitter in their party, or at least they know nothing about them. A short follow-up survey of the parliamentary party groups' (PPG) central offices on the nature of these guidelines indicates that all parties have social media style guides and that some parties coordinate their social media activities by asking MPs to follow the PPGs relevant policy issues (e.g., of their week's motions and speeches). At the same time, all PPGs confirm that MPs are ultimately free to use their social media channels as they see fit. This might explain why we, to different degrees, observe contradictory responses within the parties (AfD excluded) in our sample.

Conclusion
Do (political) elites operate social media accounts themselves, or do they assign this task to others? And are they free to behave how they want to on Twitter, or is there some sort of party control? These questions are relevant because many conceptualizations of and theories explaining social media behavior equate social media accounts with the actors they represent. And, we usually assume that the posted content reflects those actors' actual preferences and positions. Our data, collected among German parliamentarians, suggests that these assumptions are questionable to different degrees, which has several implications.
First, studies investigating (political) elites' social media use should make readers aware of this problem. Ideally, we would adapt our language and explicitly write about politicians' "accounts" rather than equate accounts with politicians themselves. Behind a single account, we may find a diverse set of actors, and by avoiding personalized language, we avoid drawing the wrong conclusions.
Second, we found that around two thirds of the accounts studied here are not operated by the corresponding politicians alone (see Figure 1, Plot 1) but rather by a combination of politicians and staff. Also, the number of people operating an account may vary (see Figure 1, Plot 2). On the one hand, politicians seem (mostly) free to communicate what they want, that is, there are no critical party guidelines/restrictions. On the other hand, it is often the staff that communicates. Hence, when we use Twitter to estimate MPs' positions (e.g., Barberá and Zeitzoff, 2018;Ceron, 2017;Ecker, 2017;Sältzer, 2022), we explicitly need to assume that the staff operating the account is communicating the MP's positions (homogeneity assumption). Ironically, that might depend on how much freedom a politician leaves to their staff to communicate the staff's ideas. We should assume that MPs regularly check their accounts and might have established guidelines on what their staff is (not) supposed to tweet, given the "viral" potential of tweets. Future studies should explore these relationships.
Third, in principle, it may be worthwhile to study the causes and consequences of the patterns discussed here, that is, the consequences of operating a Twitter account alone or with staff and how many people are involved. It seems plausible that the relationship between these variables and factors such as professionalization is complex and potentially counterintuitive, for example, in a professionalized setting, fewer specialists rather than many staffers may operate an account. For example, an account operated by several people, potentially even by experts in social media communication, should have much more success in terms of outreach and influence.
Fourth, there might be other strategies to identify how many actors operate an account beyond elite surveys that constantly suffer from low response rates. In some cases, all tweets of a politician's account are "signed," making it transparent for every single tweet, whether it was written by the politician or their team (cf. Bruns and Highfield, 2013;Grant et al., 2010). Other times, the Twitter bio indicates whether another private account only operated by the politician exists. Moreover, different authors may reveal themselves through the language they use. Other behavioral patterns may help identification as applied in the growing literature on identifying social media bots (Bastos and Mercea, 2019;Castanho Silva and Proksch, 2021b).
Our research note has limitations, and we should be careful to generalize from our study period and sample. However, to our knowledge, it is one of the first attempts to explore the above assumptions. Social media presence is becoming increasingly important across different societal Note: Data comprises a non-random sample collected from 709 members of the German Bundestag in 2020. Gray bars represent 95% confidence intervals.
spheres, not just politics. In parallel, we expect to observe an increasing grade of professionalization of social media usage. Those elites that can afford to do so might delegate their social media accounts to trained staff. Consequently, we need a more systematic discussion of how this may change the way we analyze social media data as researchers.