Justifying mobility restrictions during the COVID-19 pandemic: a test in multilevel governance

Abstract In public health crisis governance, effective communication has been shown to move people from awareness to compliance. This article examines the effectiveness of the communication strategy developed by stakeholders in the European multi-level governance during the COVID-19 pandemic. An original dataset of over 40,000 tweets from 65 actors in Switzerland, France, the UK, the EU and the UN is used to measure the timeliness, consistency and connectivity of tweets about mobility restrictions in the first phase of the pandemic. Analysis shows that the discourse surrounding mobility restrictions gradually becomes more politicised after an initial phase of high consistency and connectivity among actors. Additionally, low inter-level connectivity suggests a lack of coordination across governance levels, despite a strong consistency in the message. The study concludes that this pattern of communication could worsen the rising infodemic issue.

European integration, mobility restrictions have a particular importance, as the liberty to move freely within the EU is not only regarded as a key element of the single market, but also a pillar of political integration, sustaining the development of a common sense of identity and the realisation of the idea of a European citizenship. Therefore, effective MLG in the EU hinges as much on the adoption of adequate measures as on the way these are communicated to the wider public.
The political communication about mobility restrictions thus offers a compelling framework to measure two key aspects of the MLG response to the pandemic in Europe: (a) the level of coordination European MLG actors are able to sustain in times of crises, and (b) the resilience of shared identity and solidarity discourses in the multilevel setting when the core issue of free movement is at stake.
In this article we investigate how effective the European MLG of mobility restrictions is by focussing on the crisis communication of elite actors across the different levels of governance. More specifically, we look at three aspects of crisis communication: its timing (is communication early and active); its substantive consistency (is framing phased and coherent); and its connectivity (are actors forming into a communication network).
The importance of MLG has been underlined for crisis management in general and communication on the COVID-19 pandemic in particular (Angelici et al. 2021;Boin and Rhinard 2023;Généreux et al. 2021;George et al. 2020). This is because 'crises are moments when mutually constructed societal values are acutely threatened' . Political choices necessarily affect the legitimacy of the response and herewith 'the quality of a political system's relationship with society' (Boin and Rhinard 2023: 7). A complex multinational system like the EU must strike the right balance between coordination and responsiveness to cross-country and within-country variation (Généreux et al. 2021: 699). Based on a detailed analysis of government intervention in Italy and Spain during the COVID-19 pandemic, Angelici et al. (2021: 3) underline that 'whilst coordination across borders is required at a European-wide level to face a global pandemic, regional reactions might well [signal, authors] idiosyncratic needs' . They continue noting that 'hence, a "one-size-fits-all" approach might not always be the most efficient solution, especially when the impact of policies is highly uncertain as in the presence of a completely new virus' . Apart from yielding more efficient solutions, according to Boin and Rhinard (2023: 8) effective public communication across political levels 'can be seen as a precondition for democratic governance in the EU' and eventually also a necessary condition for 'resolving the "constraining dissensus" identified by postfunctionalists' (Hooghe and Marks 2009).
We address the effectiveness of MLG by examining communication about mobility restrictions on Twitter during the first phase of the COVID-19 pandemic. Seeking variation between EU and non-EU but closely associated European countries participating in the European free movement zone, we target governmental actors and political elites in France, Switzerland, United Kingdom, next to actors in the European Union and the UN. Linking up with the literature on effective crisis communication (e.g. Lilleker et al. 2021;Nord and Olsson 2013), we first assess the timeliness and levels of activity of actors on Twitter across phases of the crisis -from the acute phase early in 2020 to the relaxation of the first wave of measures from May to August 2020.
We then examine the consistency of the frames mobilised to justify the mobility restriction measures using claim analysis, and distinguishing between 'managerial' (i.e. technocratic, issue-focussed and health-related), 'political' (i.e. cost-benefit/economic and civic-duty-related) and 'solidarity' frames (national, European and cosmopolitan identities). Finally, we measure the connectivity of actors in terms of their interaction on Twitter and their positioning in the communication network, using network analysis. Given the interdependence among European countries in the fight against the pandemic, and in particular with regard to the regulation of international mobility, we expect that the timelier and more consistent the message, and the more connected the actors, the stronger and more operational multilevel governance is.
The analysis of a new and original dataset of over 40 000 tweets from 65 political stakeholders in Switzerland, France, the UK, the EU and the UN communicated between January and August 2020, yields mixed results on the efficiency of crisis communication in the European MLG. First, timeliness is inconsistent: after a slow start, communication picks up pace in the acute phase of the pandemic only to recede again. This uneven pattern of diffusion is known to create pockets of information void, which risk being filled by disinformation and misinformation (Purnat et al. 2021;Wilhelm et al. 2023). When it comes to framing strategies, we find variations across time and between actors: whereas the temporal evolution of framing is linked with the phasing of the crisis, ideological cleavages drive the framing distinction between actors. The latter, although signalling a functioning democracy, decreases efficiency in times of crises where a clear and consistent message helps to quell confusion and 'improve the understanding of vital information' (Eldridge et al. 2020). Finally, at the actors' network level, we report low horizontal and vertical out-group activity, meaning that the formation of networks is largely inward bound and national, despite repeated calls for international and European solidarity. State actors from all countries and across political parties focus mainly on the construction and reinforcement of domestic networks rather than reaching out towards their international counterparts.
Overall, despite the fact that frames do cross national boundaries, specifically in actors holding similar positions in the network, the European multilevel actors' network of crisis communication lacks reactivity, and remains little connected. Over time, this leads to a multiplication of messengers and suboptimal use of scarce resources rather than to the enhancement of the cooperation necessary for efficient crisis-management strategies. Moreover, as actors systematically promote national solidarity over their European and international solidarity, we find little evidence that more integration in the European MLG system is synonymous with greater participation in the supranational discursive construction. We thus conclude that the European MLG of communication about crisis mobility during the pandemic remains uncoordinated, that is to say that despite high levels of consistency, it lacks connectivity.
The remainder of the article is organised as follows. The next section reviews the literature on crisis communication and the governance of the COVID-19 pandemic in Europe. Drawing on this literature we then outline the theoretical framework guiding this study by formulating a series of propositions about the efficiency of a European multilevel crisis communication network in terms of timeliness, substantial coherence and connectivity of actors' messages. The third section presents our data and outlines our method of investigation for aggregated and topical networks at the actor and content levels. We discuss our results in a final section.

Crisis communication and the governance of the COVID-19 pandemic in Europe
Early in 2020, the COVID-19 pandemic developed into a global public health emergency in which governments and intergovernmental organisations played a central management role. From the onset of the crisis, governments and official actors have been expected to lead in all areas relevant to the pandemic, especially with regards to communication and information dissemination Edelman Trust Barometer' 2020. Because the resolution of the crisis has largely depended on change in individual behaviour, the main challenge for political elites has been to develop an effective communication strategy that will maximise compliance with new, unusual and unpopular rules, including the implementation (and relaxation) of mobility restrictions.
The literature about crisis communication shows that political elites must work in concert to form a cohesive response plan (Manoj and Baker 2007) aimed at processing and disseminating information required to address the disruptive situation in a timely manner (Coombs 2010;Fraustino et al. 2012). These findings are echoed in the Cambridge Environmental Research Consultants (CERC) framework adopted by the US Centre for Disease Control which serves as a reference in crisis communication studies . According to the CERC (2018: 8) the most common pitfalls of crisis communication are incoherent messages from multiple experts and/or government representatives; untimely and asynchronous messages across these actors; and public power struggles that may manifest themselves also through the shift from more managerial, technocratic and science-led communication to a more political one, distinguishing between the winners and losers of respective measures. Another important aspect of the social psychology of crisis communication is leaders' capacity to unite their addresses in a sense of 'we-ness' , thereby countering partisan or ideological positions and uniting the society behind the postulated measures (Jetten 2020: 25).
In order to address these issues, the crisis communication literature thus emphasises the timing of communication, the importance of different substantive frames, and the connectivity of actors. This, however, poses additional challenges for political actors operating in a multilevel governance context and using social media.

Timeliness: the challenge of multiple and decentralised actors
Such is the magnitude and scope of the COVID-19 pandemic that multiple-scale communication efforts have had to be deployed simultaneously (Ansell et al. 2010) and in a coordinated manner (Ratzan et al. 2020) to mitigate the risk of creating pockets of misinformation (WHO 2021a), information void (Noar and Austin 2020), or information overload (WHO 2021b). This is especially true in the context of a global threat where multiple agencies and levels of government are involved in the response (Simon et al. 2015). Because this pandemic reaches far beyond any local geographic zone or individual state, the network of actors producing, coordinating and disseminating the information is large, heterogeneous and has lacked the experience of communicating as a unit, thus creating hurdles to the production of a consistent coherent and coordinated communication strategy (Slaughter and Hale 2010). To add to the complexity, the crisis does not follow the same pattern in time across neighbouring countries and closely interconnected territories, making it complicated to design an effective and consistent message, that also considers regional and temporal variations (Vaughan and Tinker 2009).
The European region represents a singular subset of official actors communicating about the pandemic. Despite the comparatively higher degree of regional integration, recent research has found that the coordinated response 'started too late and ended too soon' to prove decisive in mitigating the spread of the virus (Goniewicz et al. 2020;Lichtenstein 2021). Others find that a complete lack of international coordination from the beginning of the crisis lead to competition between national responses, labelled as 'coronationalism' (Bouckaert et al. 2020). The EU itself has been shown to be a resilient actor in the face of the crisis, emerging stronger in terms of leadership (Wolff and Ladi 2020), even though its coordinative capacity has appeared uneven with some institutions largely fulfilling their mandate (e.g. the European Centre for Disease Prevention and Control -ECDC), while others have been unable to foster cooperation between member-states (e.g. Civil Protection Mechanism) (Schmidt 2020).

Consistency: synchronicity in the framing of the crisis
The substance of the political message is captured in the notion of 'frames' , understood as factual interpretations of the underlying problem and normative devices for action (Rein andSchön 1991: 264, see also Entman 1993;Vliegenthart and van Zoonen 2011). In the context of crisis communication, one can distinguish 'managerial' , that is strictly problem specific, technocratic and science led frames (in our context: health), from more political frames that emphasise either 'responsibility' (addressing the wider political and economic cost-benefit implications of health measures and or civic duties and the rule of law) or 'moral' , identity-related arguments that give contours to the sense of 'we-ness' (in our case referring to national, European and cosmopolitan identities) (Nord and Olsson 2013).
The effectiveness of frames in managing crises differs across the different phases of a crisis. According to Coman et al. (2021), in the preparatory and acute phase of the crisis, it is important that leaders rally around the science-based, managerial frame in order to give a coherent and clear message, while in the normalisation or relaxation phase, more diverse voices may enter the stage, using more political frames.

Connectivity: collaborative communication and diffusion
The challenge of aligning interests and coordinating discourses from very different actors is particularly prominent in the multilevel governance system; it is an issue that predates the COVID-19 pandemic (see Schmidt 2008;Wodak and Weiss 2005). However, in the context of the pandemic, multilevel communication faces the additional challenge of instantaneity: social media and digital platforms follow the crisis in real time, and institutional actors must invest this space swiftly at the risk of losing the public's attention (Sutton et al. 2020). In the new media environment that institutional actors are called to operate in, public health policy responses on the ground and online have become increasingly simultaneous and intertwined (Abd-Alrazaq et al. 2020).
Over the last decade, the role of social media as a communication tool during crises has gained critical importance, playing a coextensive role to traditional media (Kusumasari and Prabowo 2020). Social media even supersedes written or network media as the main news provider, 1,2 especially for younger generations. The COVID-19 pandemic has both contributed to the acceleration of this trend, and expanded the scope of actors disseminating and accessing information about the crisis online (Király et al. 2020). This is in part because social media provides an opportunity to communicate directly and in real time with the public. Social media offers the institutional actor a public platform to promote a message, rather than only a channel to report on the story. This gives governments and institutional actors more control over the narrative and to enhance their legitimacy, but also to promote more accountability over the policy results (Limaye et al. 2020;Song and Lee 2016).
In sum, combined insights from the MLG approach of political interactions and crisis communication on social media, indicate that effective communication does not simply arise in the face of shared threats: it demands strategic planning and political will. However, common interests in the face of a crisis can act as a catalyst for enhanced coordination, especially in high-stakes situations. Building on these insights, we now examine further discourse effectivity in the European MLG system. We do so by investigating the communication strategies of a network of European institutional actors during the first phase of the pandemic. We focus specifically on the discourse network about the most important and consequential decisions of the early stage of the pandemic: the implementation and relaxation of mobility restrictions.

A claim-and network-analytical approach to multilevel governance
Crisis communication literature has established the important role of the response network in information diffusion. It shows how the effective management of transboundary crises such as the COVID-19 pandemic calls for increased international collaboration among states and between different policy sectors (Ansell et al. 2010). This crisis management practice overlaps with the principles of MLG, in which the policy-making process extends beyond central governments to include vertical (political levels) and horizontal (policy types) networks of actors (Hooghe and Marks 2001). Whereas MLG literature suggests that coordination across levels of governance depends in part on the ability to communicate effectively across system boundaries (Lyall and Tait 2004), it falls short of examining the dynamics of interaction for coordinated communication. This has been acknowledged in a multi-authored call entitled 'Communication strategies and media discourses in the age of COVID-19: an urgent need for action' (Généreux et al. 2021). This concludes that 'Multi-level governance analyses … of communication strategies and information flow could also be conducted through case studies in various jurisdictions' in order to retrace actors' interaction, inter alia using network analysis (ibid.: 1183).
In this article, we assess the effectivity of crisis communication as a key element of governance systems when it involves different types of actors from multiple layers of the international governance system. More specifically, we ask: how effective is European MLG in times of crisis in terms of the contents and coordination of political communication?
We define effective communication as the ability for a given group of actors to conjointly deliver a timely and consistent message. Early and continual communication efforts lessen uncertainty, while consistency ensures that people do not have to cope with an excessive amount of information. In the context of MLG, this further implies levels of coordination that translate into systemic connectivity between different levels and types of decision-makers. We conceptualise efficient communication in terms of: 1. the timeliness and synchronicity of communication across the phases of the crisis; 2. the consistency in terms of the frames mobilised to justify mobility restriction/relaxation measures. We distinguish between: 'managerial' (i.e. technocratic, issue-focussed and health-related), 'political' (i.e. cost-benefit/economic and civic-duty-related) and 'identitarian' frames (referring to national v. European v. cosmopolitan solidarity); 3. the connectivity of actors in terms of their interaction and their positioning in the communication network. Table 1 shows the crossing of consistency of messages with the connectivity across messengers leads us to four ideal-typical constellations of the quality of governance communication in a multilevel perspective: In a functioning MLG, communication is both consistent and connected: actors not only agree on what actions need to be taken and how to justify these actions, but they also present high levels of interaction, reinforcing each other's messages by creating efficient chains of information diffusion through multiple and repeated connections. Overall, this lowers uncertainties and increases the legitimacy of the message. Conversely, if one of these two aspects is lacking, the effectivity of the system decreases. In a highly connected but inconsistent scenario, we find a disjointed system where actors, while keeping a high level of interactions, use different frames emphasising opposing courses of action. This disjointed system is more conflictual and confusing and less effective. It leads citizens to 'pick a side' and recreates cleavages and politicisation where a coordinated health policy would demand convergence and unity.
In a consistent but unconnected model, which we refer to as uncoordinated, discourses largely converge but remain parallel to one another. The multiplication of the same discourse coming from different sources is a waste of time and resources at a moment where these assets are in short supply. This can lead to information overload for populations, and overburden messengers who must control all aspects of the communication, whereas better cooperation at the systemic level would allow a more strategic distribution of communication tasks. In such distributions, the most trusted messengers could be tasked with diffusing the most sensitive information: in the case of the COVID-19 pandemic, research has found that these trusted actors are scientists, health officials and doctors (Edelman 2020). Finally, absent connectivity and consistency between actors means cooperation in crisis communication is minimal and scattered. In this 'no MLG' scenario, we cannot speak about any significant level of MLG of the crisis communication. Rather, we observe a multiplication of actors, each reacting individually and without attempting to coordinate actions or discourses.
In order to determine the effectivity of MLG in the political communication of mobility restrictions during the pandemic, we perform a comparative discourse and network analysis of positions about mobility restrictions in tweets by political elites in France, Switzerland and the UK, including EU and international actors, during the first wave of the pandemic. We selected these three countries because while they all participate in some respect in the European free movement area, they differ in their degree of integration in the EU multilevel system: France is a full member-state, the UK has just left the EU -and, as an insular country, has greater control over its national borders -and Switzerland is an associated country, surrounded by EU members. This will allow us to test whether variations in the depth of supranational integration influence the levels of cooperation between actors.
At the intersection of the crisis communication and MLG literature, we develop three hypotheses on the timeliness of the communication, and the capacity of actors across levels and types to implement consistent and connected communication strategies. First, building on recent findings on crisis communication during the COVID-19 pandemic (e.g. Bélanger and Lavenex 2021; Noar and Austin 2020), we expect a change in the effectivity of communication between the phases of the crisis, as new actors and topics beyond health issues enter the discussion: (H1) Communication effectivity in the European MLG system is higher in the acute phase of the pandemic when information is scarce and actors less partisan, and lower in the subsequent normalization phase as health issues are becoming more politicized.
Second, and in line with MLG literature, we also expect variation between countries and types of actors. Accordingly, we hypothesise that the MLG of the COVID-19 pandemic is more efficient: (H2) In France, as the most vertically integrated actor, and then it follows a downward path to Switzerland, and is lowest for the UK.
(H3) In the relations between technocratic actors (i.e. the national, EU and UN health agencies) than between governmental actors and political parties.

Research design
In order to examine these propositions, we conduct a content and network analysis of a set of COVID-19 related tweets produced by political actors in the European multilevel governance system during the first wave of the pandemic. The choice of the microblogging site Twitter is motivated by the importance this social media platform has gained in Europe and in political crisis communication more specifically . With over 86 million active users in Western Europe alone, 3 Twitter has been one of the main social media platforms used by official actors to disseminate information during the COVID-19 pandemic. On Twitter, individuals can access health information provided by trusted actors who are easily recognisable by the blue checkmark associated with their account. During the period at stake, the actors we selected posted over 80,000 tweets on the topic of COVID-19, supporting the idea that information diffusion on this social media platform is an important part of the communication strategy of stakeholders. Our objective is to determine whether political communication about mobility restrictions is consistent and connected across the network of actors because these factors influence compliance levels (Reddy and Gupta 2020). We measure consistency through the qualitative content analysis of tweets (timeliness and framing consistency) and connectivity by performing a network analysis of retweets, replies and mentions between actors in the European MLG.

Actors
In order to build our discourse and actors' network, we first identify the most prominent actors in charge of designing the strategic response to the pandemic at the highest national, European and international levels in France, the United Kingdom, Switzerland, the European Union and the United Nations. We selected actors from the executive branch (heads of state and governments; interior, exterior and health ministries), the legislative branch (all major political parties), and top agencies in charge of coordinating and implementing the new mobility directives (police and border agencies as well as health agencies). Overall, we identified 65 actors across countries, levels and types. 4

Discourses
We collected all tweets from these 65 actors for the period between 1 January and 31 August 2020. To retrieve the tweets, we used the Twitonomy 5 analytics and management tool which we cross-referenced with the Twitter standard search application programming interface (API). Our initial search consisted of a general tweet scraping for all actors in our database. We then first identified COVID related tweets by looking for tweets that included mention of 'Corona' , 'COVID' or '2019-nCov' . But because public health tweets about COVID-19 are designed to inform and convince, many tweets on the topic do not include a direct 'COVID-19' reference. To capture those tweets, we further built an extensive search dictionary from the empirical data. To create this dictionary, we first manually coded 1,000 tweets to identify the main keywords in COVID-19 tweets, which we used for a rough filtering of the data. We then manually reviewed the discarded tweets to identify whether we missed COVID-19 tweets. In the end, we still covered the whole period manually, the dictionary was a tool to help us filter in immediately the most obvious COVID-19 tweets. Using this dictionary allowed us to identify a much wider number of COVID-19 related tweets (almost double the initial number), making the dataset more comprehensive while allowing us to produce the extensive manual coding in a timely manner.
We extracted and stored the text and metadata of the tweets using the timestamp, unique ID, number of likes, retweets and user profile information including place of origin and followers. We coded tweets both in English and French and removed all tweets in other languages from the dataset. This is especially relevant for the case of Switzerland where the same tweet often exists in German and/or Italian. Overall, we identified 80,057 tweets for the period out of which 40,923 were marked as COVID related (51%).

Data preprocessing
This research is interested specifically in the subset of tweets that discuss mobility restrictions. 6 So as a preparatory step for the formal coding of tweets, we first manually coded a general category for each tweet 7 : 8,629 tweets (21% of the dataset) were identified as mobility related, in one of three categories of movement (international travel, domestic mobility and social distancing). As a specific preparation for the network analysis, we identified the source and target user for all tweets flagged as 'retweets' and 'replies' . We also extracted all mentions of other Twitter users in every tweet (identifiable by the use of the '@' symbol before the username), and similarly labelled both the source and target users for each pair of actors. As a final step, we transformed our wide data into long data to accommodate the requirements of network analysis.

Data analysis
The processed tweets were each individually coded by a team of three coders. The codebook was designed to identify in each tweet (1) the actor's position about mobility (encourages enhanced mobility v. encourages the limitation of mobility) and (2) the argument they use to justify their position, what we refer to as 'frames' . In line with recent research on the framing of the policy response to the COVID-19 pandemic ) and building on our extensive dataset of tweets, we use a mix of inductive and deductive methods to identify 6 main arguments commonly used by stakeholders in justifying their position on a given mobility restriction. For the purpose of this study, we have then aggregated these arguments into 3 overarching categories, from more technical (managerial) to more politicised (political and identitarian) arguments. Table 2 summarises the different arguments, aggregations and frequency of use by actors.
For more context, Table 3 provides examples of tweets for both positions (support the implementation of mobility restrictions measures; support the lifting of mobility restrictions measures) and each of the 6  to assist, the @ecDc_eu will provide guidance and mapping of #coronavirus transmission levels.
Utilitarian (economic costs) if the government wants people to quarantine, then they must make sure people have the financial means to do so. We need the Government to show us an exit plan. Without it, millions of people and businesses could face years of hardship after the #coronavirus #lockdown ends.

Normative (Civic duty)
let's keep going and continue to flatten the curve. Don't undo our hard work. #stayHomesavelives passengers are now able to travel to certain countries & territories without having to self-isolate on their return to england.
Be #travelaware -remember to complete the passenger locator form before returning to the uK.
Communitarian (national solidarity) pour protéger les Français et permettre à nos hôpitaux de faire face au coViD-19, la période de confinement est prolongée jusqu'au 15 avril. the government must do everything it can to bring British citizens who are stuck home. We welcome the steps the government is taking to charter flights in order to get people home, & to work to ensure commercial flights remain available where possible.
European (European solidarity) the @eppGroup strongly supports the new measures proposed by president @vonderleyen to introduce temporary foreign travel restrictions to eu. coordinated eu level measures are needed. We must protect schengen area, the supplies of goods and also citizen's health and safety. today, #schengen, one of europe's most valuable achievements is at risk! Faced with #coViD19, member states re-established internal borders in the most uncoordinated manner. How will the commission help save schengen and the principles of free movement?
Cosmopolitan (International solidarity) in response to the #coViD19 pandemic, we're providing visa extensions for those affected by #coronavirus. nobody will be penalised for circumstances outside of their control. our @DamiencareMe says we'll #leavenooneBehind in the #coViD19 crisis! eu must protect refugees & asylum seekers in Greece. We ask commission to evacuate people at high risk, increase hospitalisation capacity & speed up relocation note: all these tweets can be found in the online dataset with all the metadata. different arguments (pragmatic, normative, utilitarian, communitarian, European, cosmopolitan). Over three series of inter-coder's reliability tests, the coders maintained an average kappa score of 0.75 for the coding of four different objects per tweet: general category, mobility subcategory, position and frame. 8 We first look at the distribution of positions and frames through time to assess whether the political communication about mobility in the European MLG is timely and consistent with the dynamics of the crisis. We measure consistency through time by sequencing a time series of position and frames distribution across the period, which we compare with the number of new cases within our country cases.
We then measure actors' frame consistency by mapping and comparing the distribution of frames by actors across levels (national, European and international) and type (executive, legislative and top agency) for the two types of mobility measures (restriction and relaxation). Higher levels of consistency are associated with greater efficiency of the political communication in the network of actors.
Moving on to connectivity, we use network analysis to identify the intensity of relations between actors across levels and types. We extract the communication networks by analysing the retweets (RT), replies and mentions (@), and visualise them using the Gephi network analysis software. We use a directed network to represent the information flow between each pair of actors retweeted or mentioned, where the weighted edges between two actors represent the number of times they reference (RT or @) each other. The sender ('source') is said to have appropriated the message (RT) or included another user ('target') in the conversation (@). To determine the connectivity of the network, we look at the relations between the 65 institutional actors, in-group and cross-group. Higher vertical connectivity (between levels) suggests more coordination within the MLG network. Once again, more connectivity is associated with greater efficiency of political communication in the network.

Timeliness and sequencing
We have seen that communication efficiency reaches a peak early in the pandemic, before politicisation of the issue takes over. The urgent nature of the crisis forces the actors into reacting with a limited, but forceful number of recommendations in order to deal with the most pressing issue of protecting the population against the new threat. Then, after this 'honeymoon' phase where political decision-makers from all sides converge on the course of action, and as new decisions bring more negative socio-economic impacts, communication becomes more conflictual as we enter a phase of politicisation of the pandemic measures.
Looking at the data, a clear temporal pattern emerges when it comes to the promotion of mobility restrictions. Figure 1 shows the distribution of support (n = 4570) and opposition (n = 4060) across time, and its consistency with the cycle of increase -decrease -increase of new cases across the period. We find a significant peak of support for mobility restrictions around mid-March when the cases are rising and lockdowns are announced across our cases (March 12th for Switzerland, 13th for France and 24th for the UK). Then, we find support turning towards the relaxation of the measures as the peak of cases passes. However, this second phase is not so straightforward and from early May until the end of the period, the positions are almost equally distributed between support for keeping restrictions in place, and support for relaxing these restrictions. This mixed-message configuration shows important division and potential conflict within the network about how to move forward during the normalisation phase. It can also create confusion among the population over which behaviour to adopt as the peak of the crisis is passing. Overall, as expected by H1, we find a pattern of early rallying, followed by a polarisation of position, with no clear guidance emerging until the end of the period.
Importantly, political elites failed to increase communication levels at the start of the second wave of the pandemic which took off at the end of July. In fact, they maintained the low level of communication of the normalisation phase even though infections were starting to peak again. In terms of timeliness, this means that they were already a step behind when the second round of lockdowns hit in the autumn of 2020.
Looking further at connectivity patterns, Figure 2 shows a similar distribution of early strong connectedness between actors followed by a sharp decrease in interactions as the period progresses (n = 2778). To be most effective, communication at the network level must be timely, sequenced and active. In the European MLG network of political communication over mobility restrictions during the pandemic, we find a strong initial response by the actors, coupled with good interaction and a coherent position. However, we also observe that passed the peak, all 'unity' indicators tend to drop. Overall, the network shows more conflict over measures and less activity as the crisis unfolds, which is in line with our first hypothesis.

Vertical network: consistence and connectivity across levels
We contend that effectivity varies across actors: the more actors are integrated in the European MLG, the more consistent and connected their communication should be with the rest of the network. To test for consistency, we compare the distribution of frames supporting mobility restriction/relaxation for each actor. The more actors use a similar distribution of frame, the more consistent they are with each other. Vertical cohesion indicates effective communication across levels of governance. Horizontal cohesion points to a disconnect between levels, where actors' discourses run parallel rather than reverberate and strengthen each other.  Figure 3 illustrates the distribution of frames between actors (which frames actors use to justify their position). We find many similarities between all actors in our network. The two most used frames for all actors are health protection and solidarity. The latter also vary according to the level: countries have a majority of national solidarity frames, the EU presents more European solidarity frames, and the UN's identity frames use is dominated by international solidarity frames. This variation is nevertheless consistent with the actor's level: each actor puts forward their own version of identity and solidarity, but the principle remains the same.
In terms of MLG, however, these findings also show the predominance of national frames of belonging over European and international ones. In the context of measuring limitations of mobility and the freedom of movement in Europe, the weakness of European solidarity frames both in the messages supporting and those relaxing mobility measures corroborate the predominance of the national sense of belonging across countries' political elites. While this finding applies to all three countries, it is interesting to note that Switzerland's political elites, despite freedom of movement being strongly politicised domestically, mobilise the frame of European solidarity more frequently than France, which, according to our expectations, as a full EU member state should be the most Europeanised one in terms of MLG. At the same time, Swiss political elites emphasise much more often the national solidarity frame, whereas French elites tend to favour more neutral political frames of civic duty.
The second variable used to study effective governance is network connectivity, which we test by investigating the structure of actors' interactions on the issue of mobility. We do so by looking at the modularity of the network, which is a measure of the strength of division of a network into communities. A network with high modularity has dense connections between actors within communities, but sparse connections between actors in different communities. Figure 4 shows the network modularity for the totality of actors in our network. The nodes represent the actors: their size shows their degree of centrality (overall importance in the network) and their colour illustrates the community to which they belong. The edges are weighted according to the frequency of connection between actors, and they are coloured according to the source of the connection. The overall modularity score of the network is high at 0.712 (on a scale from 0 to 1). This indicates that communities are strongly connected inwards but scarcely connected between each other. The main component of the network is made out of 5 communities almost perfectly intersecting with our cases (UN, EU, CH, F and UK). In this network, outwards connections only make up 2% of the network (62 out of 2,840 links). The most connected cases are the UN and the EU (42 links). On the other hand, three pairs of cases are not connected at all (CH-UN; FR-UK and EU-UK).
In terms of MLG, these findings show very weak connections between the different levels of governance. This is true across all actors, so we don't find a significant variation between countries despite different levels of supranational integration. France, the most europeanised country in our sample -counts only 6 outside connections (4 towards the EU) on a total of 704, which is equally weak both vertically and horizontally as Switzerland (7 out of 219) and the UK (1 out of 852).
Our findings on vertical efficiency of the European MLG are mitigated: whereas inter-level substantial coherence is quite strong, inter-level connectivity is extremely low, even for parties which share high institutional and political integration such as France and the EU. Our findings suggest an uncoordinated vertical MLG, where similar discourses run parallel to each other, wasting the messenger's resources and burdening citizens with excessive information.

Horizontal network: consistency and connectivity across types
Our third hypothesis suggests that MLG of crisis communication is more efficient in the relations between technocratic actors -i.e. the national, EU and UN health agencies, than between governmental actors and political parties. The disaggregation of the data across different types of actors at the three levels of governance confirms this hypothesis. When it comes to consistency, and perhaps not surprisingly, technocratic actors -the health agencies -are the most consistent in their communication, and almost exclusively use the managerial frame (Table 4). Government leaders also show a relatively strong attachment to the managerial frame (Table 5), but include a much more identitarian discourse as well. Legislative actors, however, show more divergence across countries and political families, and the discourse is less concentrated around health (27% of the time, compared to 32% for heads of state and 60% for health agencies ( Table 6).
As shown in Figure 5, when comes to connectivity between actors of the same type across levels, we find a higher proportion of outreach from health agencies than both other types of actors. Health agencies cross national and organisational lines almost 20% of the time by connecting with other health agencies (40 connections out of 248 in total). The closer the agency, the stronger the connection: while the WHO, WHO_Europe and ECDC_EU share most of the connections, the other agencies remain largely outside the network. The situation is even worse among heads of the executive (4 connections between Ursula von  der Leyen and Charles Michel) and political parties (only 2 connections overall between the Modem and Renew Europe). This complete lack of connectivity is surprising because many of these actors share ideological positions and could use the reinforcement of other similar actors to push their agenda across Europe. This is especially true when such obvious bridges exist between national and European (or Europeanist) parties. As expected by hypothesis 3, the more consistent and connected actors are the non-political, institutional health agencies. But if they are strongly convergent in their discourses almost entirely centred around health protection framing, they remain comparatively and modestly connected between themselves across levels. The only exception is when there is an actor who can provide an institutional bridge between two agencies, such as the WHO_Europe for the UN and the EU.

Conclusion
Two central findings flow from this analysis of political elites' communication via the social media platform Twitter during the first wave of the COVID 19 pandemic in Europe.
Firstly, political elites from the executive and legislative branches and across levels of governance have converged in a highly managerial framing of the mobility restrictions imposed to counter the pandemic. The main argument invoked to justify both restrictions and relaxations is health protection. This provides for a strong degree of consistency, which, from the point of view of crisis communication, is seen as conducive to effective political responses. The crossing of political communication data with infection rates also shows that this political communication occurred in a relatively timely manner in the acute phase of the first wave of the pandemic. However, this timeliness decreases over time and political elites failed to anticipate the onset of the second wave starting in July 2020.
The relatively high consistency focussed on the managerial health frame is stronger for health agencies and heads of government than for political parties, but still the latter have on the whole stuck to this framing, thereby providing for a relatively low level of politicisation. Beyond health, the second most popular frame used in the political communication is arguments related to political community and solidarity, i.e. identity frames. In this regard, we find more variation across our levels of analysis. From the perspective of MLG in Europe, our findings corroborate what other scholars studying policy making during the pandemic have coined 'coronationalism' (Bouckaert et al. 2020). When it comes to identity frames, in all three countries it is the national community that serves as primary reference. Although the tweets analysed in this article focus on mobility restrictions which, in so-far as they concern national borders, imply also a major blow for the key principle of freedom of movement in the European Union, French political elites are not more 'European' when referring to political community and solidarity than their Swiss and UK counterparts. The European identity frame only prevails in the political communication of the EU actors, while the European Council President Michel shows also a strong sense of international solidarity which is otherwise only shared by UN actors.
Secondly, the relatively strong discursive consistency around the managerial health frame is not matched by our additional measure of effective MLG, actor connectivity. Overall, we find very little connectivity between actors outside their respective level or type. National actors concentrate on their domestic network and they do not use international or European platforms to boost their messages. The same is also true the other way around: European actors do not seek to penetrate the national discourse in member-states or associated states. Even the international sphere, which comprises the sum of national actors, seems uninterested in bridging the different levels of governance, despite repeated calls for solidarity. In other words, consistency of the discourse does not translate into practices of cooperation in the communication strategy.
Interaction is also limited between types of actor, that is to say political parties, agencies and heads of states. Actors do not make the quite easy step of connecting together on social media, despite ideological closeness, shared goals and plenty of potential influence into which to tap. Again, this shows that national actors do not see the gains of fostering connectivity at the European and international level. This could explain some of the lateness in reacting to the common threat at the beginning of the pandemic, or even failures to implement supranational policies (early regulation of mobility in the Schengen space for example). This is not necessarily surprising given the context within each nation as well as the relationships between the nations and bodies included in this study. Rather, it confirms MLG assumptions that established functional relations are more likely to be sustained during crises than news connectivity to be established, even in times where actors could benefit at all levels.
However, in a crisis that is simultaneously unfolding online and on the ground, this lack of connectivity on one of the most popular social media platforms across the world presents as a failed opportunity to use technological tools to foster multilevel governance and devise a comprehensive crisis communication strategy, which could go a long way to improving health outcomes. This lack of connectivity and reactivity between official actors opens the risk for disinformation and misinformation to spread quickly at the transnational level; the objective of providing accurate information in a timely manner during crises cannot be achieved without tackling this issue.