Opposition in times of crisis: COVID-19 in parliamentary debates

Abstract The COVID-19 pandemic presents an exceptional crisis situation not only for governments, but also for politicians in opposition. This article analyses opposition party expressed sentiment vis-à-vis government actions and policies during the first six months of 2020. Based on an original content analysis of parliamentary debates in four established parliamentary democracies (Germany, Israel, Netherlands, United Kingdom), relatively positive opposition expressed sentiment in parliament early on during the crisis is observed, in line with a ‘rally effect’ observed in public opinion. Sentiment turned more negative as the first wave of the crisis abated. Larger opposition parties with considerable prior government experience were more positive than larger parties without such experience.

The COVID-19 pandemic presents a crisis situation that requires extensive government action, whereas the role of parliamentary opposition parties is less clear. Government-opposition relations during a crisis can range from consensus seeking to strictly adversarial. The COVID-19 pandemic presents a special type of crisis: primarily a public health crisis, but with important social and economic consequences. It directly affects the daily life of many citizens as well as work in parliament, impacting the way in which governments can be held to account.
From a normative democratic perspective, government-opposition relations in times of crisis may entail contrasting implications. On the one hand, government-opposition consensus on crisis measures increases their legitimacy among the public. On the other hand, the opposition's criticism is important as a catalyst for public discourse, which is vital for democracy especially in times of crisis when governments are prone to seek more powers and to weaken checks. It is therefore important to understand how opposition parties behaved during the first months of the COVID-19 crisis.
We analyse opposition parties' parliamentary behaviour, more specifically their expressed sentiment vis-a-vis the government in parliamentary speeches, during the first six months of 2020 and explore explanations for variation in this sentiment over time and between opposition parties. Building on the 'rally around the flag' literature, our main expectation is that as the pandemic develops, and the nature of the crisis changes from mostly an acute public health crisis to a mix of public health, social and economic crisis, opposition parties' expressed sentiment towards the government in parliamentary debates on COVID-19 shifts. We expect expressed sentiment to follow a curvilinear pattern over time, starting out at regular, moderately negative levels, subsequently becoming more positive as the COVID-19 crisis reached a high point, and then becoming more negative as the most urgent and severe stage of the public health crisis passed. In addition to this general trend, however, we expect general differences among opposition parties to persist, with ideology, party size and government experience playing a part in determining expressed sentiment.
We employ a comparative case study approach, selecting a diverse set of established parliamentary democracies in terms of their political institutions and the crisis' overall severity: Israel, Germany, the Netherlands and the United Kingdom. We measure opposition party expressed sentiment towards the government using original manual content analysis of parliamentary debates on COVID-19 related topics. A pooled analysis of the data explores the explanatory power of crisis-related factors and opposition party characteristics. The data show a monotonous trend towards more negative sentiment over the course of the pandemic (up until early July 2020) that provides some support to a rally-around-theflag dynamic. In line with previous research, we also find more positive sentiment among larger opposition parties with more prior government experience, while larger opposition parties without prior government experience are more negative.

Opposition behaviour in times of crisis
Legislative conflict in parliamentary democracies is usually best understood in terms of the distinction between opposition parties on the one hand and government and governing parties on the other hand (King 1976). Still, the relationship between government and opposition rarely is purely conflictual: most opposition parties support at least some government legislation or show other forms of cooperative behaviour towards the government (Andeweg 2013).
In order to understand the impact of the COVID-19 crisis on opposition behaviour, it is helpful to consider the effects of crises on government-opposition relations more generally. The most far-reaching option is for the opposition to join the government, of which the Wartime governments in the United Kingdom are prime examples (Fairlie 1918;Jefferys 1991). As the COVID-19 crisis is perceived as a national emergency, albeit of a different nature, the crisis has similarly been used as an argument in government formation, especially in countries that did not have a fully-fledged government at the moment of crisis onset. Beside Israel, which we will discuss below, this occurred in Ireland, where Fianna F ail and Fine Gael agreed to enter coalition talks for the first time ever, citing COVID-19 as a propellant (Leahy and Kelly 2020). In other cases, the opposition was invited to join the government's efforts to deal with the crisis from the outside. In Belgium, the incumbent minority caretaker government reached a deal with opposition parties for supporting crisis measures, and in the Netherlands an opposition party member and former junior health minister temporarily stepped in as Health minister when his predecessor resigned for health reasons.
Such agreements and appointments are rather exceptional, but we expect the relationship between government and opposition to be affected by the crisis in other parliamentary democracies as well. A public opinion 'rally around the flag' is a common pattern during major international crises, such as the start of a war or large-scale terrorist attack. Such patterns have been shown to arise in political systems as disparate as the United States, Great Britain and Israel (Baum 2002;Brody and Shapiro 1989;Feinstein 2018;Hetherington and Nelson 2003;Lai and Reiter 2005;Mueller 1970). During such an event, public support for governments often, but certainly not always, increases (Mueller 1973). While the 'rally around the flag' argument was developed in the context of public opinion, Chowanietz (2011) shows that there is a similar pattern of elite rallying in times of crisis: opposition parties' criticism of the government markedly declines after acts of terrorism.
Two explanations have been offered for such an increase in public or elite support. The patriotism school argues that this stems from a feeling of national unity, the idea that despite any differences that might exist, in times of crisis we need to come together to support 'our side'. Others point at the importance of opinion leadership (Brody and Shapiro 1989), noting that rallies do not occur in all crisis situations, but only when opposition opinion leaders refrain from openly criticising the president or government. During a rapidly developing crisis, the information available to governments is far superior to those of opposition leaders. Thus, opposition leaders 'have almost no incentive to criticise the president when the news of the crisis first breaks' (Brody and Shapiro 1989, 355). Given such lack of dissent, public support is expected to increase.
The COVID-19 pandemic is a different type of event than those usually studied in the rally around the flag literature. Nevertheless, it shares certain characteristics with a military crisis: it hit quite suddenly, with a large effect on everyone's daily life. Indeed, wartime analogies were often used in political messages about this crisis. The COVID-19 pandemic directly impacted on the way parliament could conduct its business, which is likely to have further increased politicians' perceptions of the pandemic being an external threat (Jenny and M€ uller 2020). Although the 'adversary' in the COVID-19 crisis is not as clearly defined as during a military or terrorist attack, an understanding of the immediacy and scale of the public health threat may cause a similar increase in public and opposition support for the government's actions.
In other respects, the COVID-19 public health crisis resembles economic crises having a profound societal impact but lacking a clearly defined 'adversary'. In contrast to military crises, previous work shows increased negative opposition behaviour during economic crises. The analysis of opposition behaviour during the economic crisis in southern Europe in the late 2000s and early 2010s revealed a strong decrease in parliamentary consensus, especially among non-mainstream opposition parties (De Giorgi and Moury 2015; but see Marangoni and Verzichelli 2015). It is, however, important to bear the time frame in mind: the financial-economic and Euro crisis unfolded over the period of almost a decade. In most EU countries trust in government did not start declining immediately; in some countries it even spiked in the first months after the start of the financial crisis (Larsen et al. 2019;Roth et al. 2013, 23). There is good reason to expect even stronger public support and elite consensus in the early stages of a pandemic crisis: whereas economic crises or terrorist attacks relate to salient political conflict lines in most political systems, which can be readily politicised, this cannot be said for the public management of pandemics.
Thus, we expect that the effect of the COVID-19 crisis on opposition party expressed sentiment followed the rally around the flag logic, at least in the short term. COVID-19 was an immediate, severe, and rapidly developing crisis that left the government with a huge information advantage compared to opposition parties with regard to an issue that was not politicised before. We, therefore, expect to observe increasingly positive opposition party sentiment towards the government as the COVID-19 crisis reached its initial zenith. As the urgency and immediacy of the crisis dissipated and the social and economic consequences of the crisis became clearer, we expect that this positive sentiment gradually declined.

Explaining crisis relations: crisis-level and party-level explanations
Our primary aim is to describe patterns in opposition expressed sentiment as the crisis developed. Given the current stage of the crisis and the available theories, any analysis of the variation in these patterns is necessarily exploratory. Nevertheless, several potential explanations can be derived from existing work on government-opposition relations. In line with the above argument, we will focus on the development of the crisis over time. Additionally, we will analyse the heterogeneity of responses among opposition parties.
We are aware of the impact of political system characteristics on opposition behaviour, which is more adversarial in majoritarian than in consensus democracies (Norton 2008). In majoritarian democracies political competition is closed and the main opposition party tries to replace the government party (Mair 1996). In consensus democracies multi-party competition generally offers at least some room for opposition parties to be constructively engaged in policy making and the prospect of future cooperation in government (Louwerse et al. 2017;Strøm 1990;Tuttnauer 2018). In federal states, due to crisis management responsibilities often being assigned to the federal and state levels, we would expect more cooperative behaviour as some parties are in opposition on the federal but in government on the state level (Hohendorf et al. 2021;Lijphart 2012). We will therefore analyse the below patterns in a diverse set of countries in terms of their political system. If we find that the within-country patterns hold in this diverse set of countries, it is more likely that we can generalise these findings to a broader set of parliamentary democracies.

Explanations relating to the crisis
A first set of factors that potentially impact on opposition party expressed sentiment relate to characteristics of the crisis, particularly its severity and its temporal development. The severity of the crisis varies between countries, even though comparing crisis impact is difficult, given differences in reporting of cases and fatalities between countries. Analyses suggest that in some countries the death toll in terms of excess death was relatively low in the first six months of 2020, while it was considerable in others.
Obviously, case fatalities also vary over time, and the crisis might be perceived as more urgent when case and fatality numbers are on the rise in a country or region. As a result, opposition parties could adjust their behaviour towards the government: more cooperative and positive at the 'height' of the crisis, more negative when the crisis threat is perceived to be lower.
Rather than looking at crisis severity in terms of the number of cases and fatalities, we can also think about the crisis in terms of its development over time. After all, even in countries with limited case numbers and fatalities, the impact of the COVID-19 crisis has been extensive (Hale et al. 2020). We expect opposition expressed sentiment to start at regular levels before the virus spread substantially within a country and when threat perception was still low. As the first wave of the COVID-19 crisis reached its zenith, we would expect more positive opposition party expressed sentiment towards the government, also due to the large information asymmetries between government and opposition at this time. After this shock of rapidly increasing cases and government lockdown measures started to dissipate, we would expect to see more negative opposition sentiment as there was more time to consider government actions and more information about these choices became available to opposition parties. We thus expect a curvilinear pattern of government-opposition relations over time.
A third way to capture crisis severity is via restrictive measures taken by the government such as school closures and lockdowns (Hale et al. 2020). These measures likely contributed to and reflected the feeling of urgency among the public and elites alike, even in countries with relatively low case numbers and fatalities. This factor also taps into the government's response to the crisis, with some observers arguing that taking restrictive measures early on reflected proper government action, which opposition parties would be more likely to support. Therefore, we expect to see a correlation between the stringency of government measures and opposition party expressed sentiment.

Party level-explanations
Studies of government-opposition relations in 'normal' times point to characteristics of opposition parties as relevant explanatory variables: some opposition parties take a more adversarial stance than others do. This raises the question whether such party differences persevere during the COVID-19 crisis. One could expect opposition party expressed sentiment to be uniformly positive in case of high-impact, short crises, such as terrorist attacks (see, e.g. Chowanietz 2011). When crises last longer and relate to existing lines of conflict, such as the Eurocrisis, the reverse can be observed: in Portugal, the Eurocrisis resulted in a stronger distinction between 'radical' opposition parties who became more negative and 'moderate' parties who became more positive towards the government (De . We argued above that the COVID-19 crisis sits somewhere in between. Therefore, our tentative expectation is that party differences play a similar role in shaping opposition sentiment during COVID-19 as during 'normal' times. Our analysis includes three variables that have been demonstrated to explain variation in opposition behaviour. First, larger ideological distance between a party and the government reduces voting with the government (Hohendorf et al. 2021;Tuttnauer 2018). Even when politicisation might be low during a major public health crisis, we expect that a larger ideological distance to the government leads to more negative opposition expressed sentiment.
Second, opposition parties with more government experience are more likelyabove a certain size thresholdto support government legislation (Tuttnauer 2018). In times of crisis, this 'responsible' (Mair 2009) behaviour might be even stronger than under normal circumstances (De Giorgi and Moury 2015, 118).
Third, an opposition party may be less likely to support government policy if it is large enough to achieve a position of power itself by replacing the incumbent government after the next elections (Tuttnauer 2018). Larger parties might show lower levels of support for government policies and actions because their credibility as an electoral alternative suffers from being overly supportive of the very government they seek to replace.
We expect the effects of experience and size to interact. While a small party with no government experience might be a negligible or niche political player, a large party with similarly no government experience is more likely to play the role of the outsider, challenging the existing party 'cartel' of mainstream government alternatives. The latter kind is expected to express more negative sentiments towards the government than the former. However, the difference between small and large parties may disappear, or even take the reverse direction, when government experience is considerable. In that case, it is the large party that might be especially constrained by the norms and the expectations from 'The Opposition' (Norton 2008) to rally around the political mainstream in times of crisis, and is thereby expected to express itself more positively.

Case selection
Our analysis covers parliamentary debates in four parliamentary democracies: Germany, Israel, the Netherlands and the United Kingdom. The selected countries provide theoretically crucial variation with regard to two system-level variables: the type of democracy (consensus versus majoritarian) and the vertical division of power (federal vs. centralised states). Israel and the Netherlands are positioned towards the consensus pole of Lijphart's (2012) executive-parties dimension, Germany is also positioned in the 'consensus' group, while the United Kingdom is clearly majoritarian. German is a federal state, while Israel and the Netherlands are unitary. In the United Kingdom's asymmetrical federalist system, health policy is a devolved issue, which implies that on this issue the vertical organisation of the state plays a role.
The selected countries also vary in terms of the severity with which the crisis hit. The United Kingdom and the Netherlands witnessed higher numbers of fatalities than Germany and Israel in the first half of 2020. According to The Economist Britain counted almost 65,000 excess deaths since the country's first 50 covid-19 related deaths until mid-July 2020 (The Economist 2020). This amounts to 98 excess deaths per 100,000 people. The excess death figure for the Netherlands was 55, while for Germany it was only 9 per 100,000 people. In Israel mortality levels in May were only slightly higher than normal (Wu et al. 2020).

Measuring opposition party expressed sentiment towards the government
Most studies of opposition behaviour in parliament focus on legislative voting (De Giorgi and Marangoni 2015;Dewan and Spirling 2011;Hix and Noury 2016;Hohendorf et al. 2020Hohendorf et al. , 2021Louwerse et al. 2017;Tuttnauer 2018), amendments or opposition bills (Loxbo and Sj€ olin 2017). The study of voting on bills, amendments and other proposals is, however, not very well suited for our current purposes due to low number of legislative votes during the early stages of the COVID-19 crisis. For example, between March 10 and May 12, 2020, the UK House of Commons did not vote at all.
In contrast, all four parliaments debated the COVID-19 crisis on a regular basis. Parliamentary speech has been used previously to map government-opposition relations using various methods of quantitative text analysis (Curini et al. 2020;Lauderdale and Herzog 2016;Proksch et al. 2019). Curini et al. (2020) and Lauderdale and Herzog (2016) use Wordfish and Wordshoal scaling of parliamentary debates respectively, which is quite an indirect measure of government-opposition dynamics. Proksch et al. (2019) directly measure the sentiment of government and opposition parties using a dictionary-based sentiment analysis. Opposition parties use a larger share of negative words than governing parties, which can be taken as an indicator for opposition behaviour vis-a-vis the government's policies and actions. Our concern with the application of quantitative sentiment analysis during a crisis is that the perceived crisis threat could lead to parliamentary actors choosing more negative words, especially when the pandemic is expanding. Any usage of positive words by opposition parties towards government policy and action might thus be offset by negative words regarding the severity of the crisis.
For these reasons, we use a manual content analysis of opposition speeches. We focus on debates that deal with the direct response to the public health crisis (e.g. dealing with lockdowns) or the general government strategy for dealing with its fundamental effects (e.g. a major economic recovery program or a supplementary budget). We include plenary debates from the start of the pandemic until early July 2020 (see Online appendix A for a full list). The number of days on which COVID-19 related debates took place ranges from 6 to 31 in the four countries.
Conceptually, we aim to capture the degree of manifest support or criticism that an opposition party states towards specific government action and government policy dealing with the COVID-19 pandemic. We coded speeches by opposition MPs 1 on the level of paragraphs, which are usually the units within a speech that develop a single argument. When a paragraph consisted of clearly separate arguments that merit different codes, we split it.
Each paragraph was coded on a five-point scale of opposition party expressed sentiment, ranging from 'very negative' to 'very positive' (see Online appendix B for full coding instructions). Coding proceeded in two steps. First, the coder decided whether the paragraph was negative, neutral, positive or whether it did not concern government action/policy (in which case it was coded as 'not applicable'). Neutral statements mostly included questions about government policy or actions that did not convey an opinion on behalf of the speaker. Second, the coder distinguished qualitatively between 'very' and 'moderately' positive or negative statements. Very negative statements are those that go beyond what is deemed normal criticism of the government, for example using blunt, severe language or personal criticism. Very positive statements include praise or very strong credit for the government's actions or policies, rather than simply expressing support. These finer distinctions depend on what is considered 'normal' behaviour in a specific parliament and are thus context-dependent, which we argue is a characteristic of language use. As our main inferences are drawn within countries, this does not pose a problem for our analysis. Inter-coder reliability tests show satisfactory results with Krippendorff's alpha ranging from 0.77 to 0.87 (see Online appendix C).
The codes for individual paragraphs were subsequently aggregated for all speakers who spoke on behalf of their party 2 on a specific day by taking the mean value across all paragraphs.
An analysis of COVID-19 debates alone cannot show whether opposition parties expressed more positive or more negative sentiment in these debates that they usually do. We can establish a 'baseline' for comparison by analysing a major political debate in pre-COVID-19 times. We select the most recent main 'general' political debate before the onset of the pandemic in the United Kingdom (Debate on the Address, December 2019), the Netherlands (Algemene Beschouwingen, September 2019) and Germany (General Debate on the Chancellor's Budget, September 2019). For Israel, such an analysis is impossible, because elections were held in early March 2020, which means that there is no pre-COVID-19 debate under the same parliament and government as the debates during the COVID-19 crisis. These debates are coded using the same opposition expressed sentiment scale.

Operationalisation of the independent variables
The crisis-related variables include the number of newly reported cases and fatalities (per 1 million inhabitants), obtained from Johns Hopkins University's COVID-19 Data Repository (Dong et al. 2020). While confirmed cases and fatalities are underreported in most countries, this is the information about the severity of the crisis that was available to legislators at the time. We study the temporal development by including the (orthogonalized) debate date and its squared term as explanatory variables. We rely on the Oxford Government Response Tracker to measure the policy stringency of the government measures (Hale et al. 2020). The Stringency Index ranges from 0 (no measures) to 100 (full restrictions on all included sub-items).
Our analysis includes three party-level variables. The left-right distance between an opposition party and the government is calculated from the Chapel Hill Expert Survey (CHES) 2019 (Bakker et al. 2020). We calculate the government's position as the seat-share weighted average position of the coalition parties and calculate the absolute distance to this position for each opposition party. For Israel, which is not included in the CHES, we rely on a separate expert survey that we administered in July 2020, including the same general left-right question as the CHES. Government experience is measured as the number of years the party has been in government since 1990. Party size is measured as the party's seat share in parliament. Government experience and party size are calculated using the ParlGov database (D€ oring and Manow 2018).

Methods of data analysis
Our data analysis includes a descriptive analysis of opposition expressed sentiment as well as pooled regression models to capture the impact of the crisis-and party-level explanatory variables. For the latter we apply Ordinary Least Squares (OLS) with country fixed effects and clustered standard errors by party. We are careful not to read too much into the country differences in our analysis, as the differences found might be due to differences in debate selection or coding decisions between countries. As a robustness test, we applied a multilevel linear regression model that uses the individual speech acts as the unit of analysis; this model corroborates the findings presented here (see Online appendix E).

Results
Opposition expressed sentiment towards the government became increasingly negative in each of the four countries as the COVID-19 crisis developed. Figure 1 presents the mean expressed sentiment on each day. The decreasing pattern is most clear for Germany, while for other countries there is more variation around an otherwise clearly negative trend. The estimates for the United Kingdom and Israel include a number of days with relatively short debates, which explains the larger variation.
Mean opposition party expressed sentiment is clearly more positive during the COVID-19 debates in March and April 2020, compared to our pre-COVID-19 baselines of À0.90 in the UK, À0.82 for Germany and À0.74 for the Netherlands. Opposition expressed sentiment in the later COVID-19 debates is generally much closer to those baseline levels. With the partial exception of Germany, there is little evidence of a curvilinear trend. Opposition expressed sentiment starts at relatively positive levels and does not clearly 'peak' at the height of the crisis (mid/late-March to mid/late-April), but rather declines in a monotonous fashion. However, these aggregate patterns hide relevant variation among opposition parties in each country. Therefore, we will briefly describe and discuss the patterns in each country, before turning to the pooled regression analysis. Figure 2 shows the position that the four German opposition parties took towards government actions and policies in all general COVID-19 related policy debates in the German Bundestag and two debates on supplementary budgets to remedy the direct financial implications of the crisis. The graph reveals consistent party differences. Overall, the Greens displayed the most positive sentiment vis-a-vis the government (e.g. on March 3: 'From our perspective, the government is currently doing many things right'), whereas the right-wing populist AfD was most negative according to our index (e.g. on April 3: 'Finally restore the constitutional order in this country and give the freedom back to the people!'). The scores for the Left Party and the liberal FDP are somewhat in the middle, especially when we account for the fact that the positive outlier for the Left on February 12 is based on only two coded paragraphs. This ordering of parties is in line with the competitive behaviour of the various opposition parties in recent years (Hohendorf et al. 2020;Stecker 2018). With 15 years in opposition and strong performance in current polls, the Greens are eager to get back into power and are very accommodating towards the government. By contrast, the AfD is clearly a pariah party and pursues a strictly confrontational strategy in the Bundestag. The Liberals and the Left oscillate between competition and cooperation with potential future coalition partners, also due to intra-party disputes about future strategy. In Germany, we do observe the theoretically expected curvilinear temporal dynamics. All opposition parties acted more positively towards the government when the situation was deteriorating rapidly on March 25. On that day, all parties except the AfD emphasised the constructive collaboration between government and opposition using typical 'rally around the flag' rhetoric. For example, Christian Lindner (FDP) stated that 'In times like this, government and opposition have a joint responsibility for the state'. Ralph Brinkhaus of the governing party CDU in turn thanked the opposition for lifting procedural deadlines and 'saying now it is more important that we get solutions for this country'. We also see relatively positive sentiment when the crisis was becoming more threatening on March 4 and when improvements had just begun and were quite uncertain on April 23.

Germany
One core reason for these highly cooperative opposition party sentiments is federalism. In Germany, many jurisdictions relevant for dealing with the COVID-19 pandemic lie at the state level (e.g. on many public health and general lockdown measures) so that the federal government had to coordinate its crisis reaction with the state governments (Hegele and Schnabel 2021). As Greens, FDP and Left are all members of some state governments, all parties except the AfD were part of an informal coordination regime that decided (implicitly) by unanimity and thus were de facto veto players for many important crisis measures. This joint responsibility put limits on the degree to which most opposition parties could credibly blame the federal government for its immediate crisis reaction. A second reason may be the common perception in the public that the governments on both the national and subnational level handled the situation well leading to low numbers of cases and casualties compared to the situation in other countries.
By the end of May, when the acute crises had subsided and the immediate need for coordination had decreased, opposition parties reverted to more negative sentiment. While the Greens still openly agreed with some government measures, all opposition parties increasingly voiced criticism, especially with regard to the planned economic recovery measures.

Israel
When the COVID-19 crisis hit its shores, Israel was already experiencing a dramatic political crisis. The first patient in Israel was positively diagnosed for COVID-19 on February 21 (Maor et al. 2020). Less than two weeks later, elections were held for the third time in a year, after repeated stalemates between the parties supporting the incumbent prime minister Netanyahu and those opposing him. By the time the new parliament convened, the caretaker government had already shut down the education system and non-essential businesses, and COVID-related debates revolved mainly around economic measures and the government's employment of the secret service to track infection spread.
With the stalemate between the pro-and anti-Netanyahu blocs still in place, even the question of which parties constitute the opposition was up in the air. Initially, it was the leader of the anti-Netanyahu bloc, Blue-White (BW) leader Benny Gantz, who received the role of formateur. However, on March 26, Gantz decided to join Netanyahu in an 'emergency' grand coalition, citing the pandemic as one of several reasons. This decision shattered the party system, resulting in both BW and the left Labour-Gesher-Meretz merger to split, with some factions joining the grand coalition and others, as well as Netanyahu's long-time rightwing partner, Yemina, staying in opposition. Figure 3 shows the positions taken by the four parties that were in opposition throughout the whole period. The four parties are similar in that none has been overwhelmingly positive towards the government, even early on. The first two months of the crisis were characterised by erratic responses of all four parties, registering a positive score on at least one occasion. Consequently, all parties became more consistently negative, with no such overall positive position from May onwards.
Surprisingly, the most supportive party was initially the Joint List (JL), a mostly-Arab party and the one closest to a pariah party in Israel. Its leaders repeatedly used the plenum to express solidarity in the face of the pandemic and to reach out to their voters urging them to abide by the Ministry of Health's orders (e.g. Ahmad Tibi, April 1). In line with the pragmatism and policy-driven focus characterising the Israeli opposition in general (Tuttnauer 2020), the Joint List readily supported (as did other opposition parties) relevant legislation as the crisis unfolded. This support was sometimes crucial to the legislation's success. Eager to taunt their right-wing foes and perhaps in a push towards the mainstream, JL repeatedly claimed credit for that support. At the same time, Joint List leaders vehemently attacked the government, e.g. claiming 'they will lie to your face' and calling for public demonstrations (Ayman Odeh, April 6).
For the right-wing Israel Our Home (IOH) and the left-wing Meretz, positive attitudes early on can be explained by the unique situation preceding the grand coalition's investiture, in which these parties held important parliamentary committee roles. For example, Oded Forer's (IOH) comment that '[we] will not be the brake to stop the government [from fulfilling their plans] on necessary issues regarding the Corona' (April 6) was made from a position of power, holding the role of Finance Committee Chair. The two parties differ, however, in their proclivity for extreme negativity. While Meretz was much farther ideologically from both governments than IOH, it mostly voiced moderate negativity aimed at the bills and measures. In contrast, IOH was much quicker to voice extremely negative views, in accordance with their tendency for populist rhetoric, as when Eli Avidar called Netanyahu a 'dictator', said 'this government has no desire to solve the Corona problem' and called for mass demonstrations (June 22).
The only party remaining steadily negative throughout the period was Yesh Atid (YA), the largest opposition party for which resistance to Netanyahu and their once-allies BW was visceral and more personal than ideological. While criticising the insufficient aid to businesses, the government's reliance on the secret service to track infections, and more, YA's fiercest criticism was directed towards the grand coalition's formation itself: 'People have nothing to eat, businesses are collapsing, in hospitals people are dying … In these circumstances, forming a bloated corruption government is a national disaster.' (Yair Lapid, April 16).

The Netherlands
Dutch opposition parties' expressed sentiment towards the government's policies and actions shows clear variety between parties and over time (see Figure 4). Some opposition parties were quite critical towards the government throughout the period of analysis, most clearly the Freedom Party (PVV) and to a somewhat lesser degree the Socialist Party (SP). These parties are ideologically far removed from the government. Other parties were more positively inclined from the start and mostly remained that way, particularly the small orthodox protestant SGP and the 50PLUS party. Despite its perennial opposition status, the SGP has had a 'governmental' attitude for a long time. 50PLUS was very concerned about the impact of the crisis on the elderly and addressed this concern constructively in the debates. A move from more positive to more negative stances is visible for most of the other parties, most clearly so for GreenLeft (GroenLinks). This party started out with positive index scores ('I greatly appreciate the way in which the government attempts to face up to the crisis') which gradually transformed in more a negative appreciation of the government's policies and actions ('This is a useless arrangement.'). This negative trend is also visibleto a slightly lesser degreefor Labour (PvdA) and DENK. The Party for the Animals (PvdD) was initially quite negative, partly related to the source of the new virus, which they link to the way in which animals are kept, subsequently more positive during the height of the crisis and again more negative afterwards. Forum for Democracy (FvD), a small right-wing populist party, alternated between being critical of a lack of lockdown measures (mid-March) to a more positive stance when such measures were taken, gradually moving towards more negative positions in May and June.
Perhaps surprisingly, none of the Dutch parties score very positively on our opposition expressed sentiment scale, not even at the zenith of the crisis. This can be partially attributed to ideologically distant parties without government experience, such as PVV, SP and FvD, all of which can beto a varying degreeregarded as populist parties (Meijers and Zaslove 2021, 22); these parties displayed particularly negative sentiment. An additional explanation is that support for the government's policies was oftentimes implicit: most opposition parties did not spend many words supporting or praising the government, but they also did not oppose government policy. This is not uncommon, particularly on issues that are not highly divisive or politically salient.
Contrary to the German case, Dutch opposition parties were not directly involved in COVID-19 policy making either. The government made the decisions based on advice from experts; opposition parties were not directly involved in the policy-making process. In the parliamentary debates government ministers showed a willingness to take into account opposition parties' concerns, but this was after policy was initially set by the government. In that sense the old Dutch political 'rule of the game' that the government has the right to govern with little influence from parliament was certainly revived in this case (Lijphart 1968).

United Kingdom
'I appreciate the cross-party approach that is being taken to this outbreak as reflected in the shadow Minister's remarks', commented Secretary of State for Health, Matt Hancock (Conservative) in the very first debate on COVID in the House of Commons, on January 23, 2020. The consensual tone of that debate, especially by British standards, is striking as it was more than a week before the first two confirmed cases in the UK, with speakers expressing hope that the virus would not reach it.
This less adversarial style cannot be explained by the policy area concerned: the National Health Service (NHS) is frequently the topic of fierce disagreement in British politics. Now, such disagreement was referred to only obliquely. On February 3, Jonathan Ashworth (Labour), the Shadow Secretary of State for Health, commented, for example, that 'This is a time of considerable strain on the NHS. I know the Secretary of State and I disagree on why this is, but he will accept that it is a time of high pressure.' (italics added). Figure 5 shows the average scores for each of the parties over time. The Democratic Unionist Party (DUP) has been most supportive of the government throughout this period: 'I congratulate the Secretary of State on his clear leadership on this matter and his determination to deal with the issues' (Jim Shannon, DUP spokesperson for health, February 26). The DUP is ideologically closest to the governing Conservative party. The two largest opposition parties, Labour and the Scottish Nationalist Party (SNP) showed very similar patterns of opposition expressed sentiment, although the content of their interventions differed: SNP spokespersons would often express the need for all countries within the UK to coordinate their responses (health policy is devolved to the Scottish, Welsh, and Northern Ireland governments), while Labour spokespersons would often criticise the government's 'neglect' of the crisis' impact on the poor and on minorities.
That the smaller opposition parties seem to have taken a more negative position may be an artefact of their spokespersons facing more time constraints and usually speaking later in debates. This may have prompted these parties to prioritise their questions and criticisms rather than repeat statements of support for the government's measures made by previous speakers. The gradual erosion of cross-party support for the government's approach to the crisis ismore or less clearlyvisible for all opposition parties. One might have expected an impact of the leadership change in the Labour party in early April, but Keir Starmer's promise of a more constructive opposition generally does not seem to have delayed his party's stepwise return to more adversarial behaviour with regard to COVID policy. It is important to note that this shift among opposition parties did not wait until the heat of the crisis was over. This may be related to the fact that while most of the early debates dealt with the public health response to COVID, gradually other policy domains became involved as well, such as the package to support businesses and prevent mass unemployment, and the disproportional impact of the pandemic on the BAME (Black, Asian and Minority Ethnic) community. On such issues, ideological distinctions may have prompted more negative rhetoric even early into the crisis: 'I am afraid that it is quite evident that the Government have lost support and confidence across the Chamber, and that is echoed by the markets.' (Matt Western,Labour Opposition Whip,March 19).
The government's crisis management also provided occasion for opposition parties to take a more negative stance: for example, prime minister Johnson's May 10 announcement of an easing of lockdown, without first consulting the leaders of the devolved governments or alleged breaches of lockdown rules by the prime minister's chief adviser, Dominic Cummings, in April (see also Fancourt et al. 2020).

Pooled regression analysis
The results of our pooled analysis are presented in Table 1. We first estimate a baseline model consisting of only the date and its squared value (orthogonalized and transformed so the date variable ranges between 0 and 1) along with country fixed-effects. This model confirms that the effect of passage of time was monotonously negative, and not curvilinear, since the date-squared coefficient is insignificant.
Model 2 replaces the date variables with our crisis-level variables. We observe no significant relation between case numbers, deaths and opposition expressed sentiment. Contrary to our expectation, the relationship between the stringency index and opposition sentiment is negative. This can be explained by the fact that stringency was low in early March, but expressed sentiment was relatively positive, while later on in the crisis stringency was high, and opposition expressed sentiment negative. The overall fit of model 2 is worse than that of model 1, suggesting that opposition expressed sentiment is not just a response to objective criteria of crisis severity, but that timing matters. Opposition parties were willing to 'cut the government some slack' early on, when the crisis first hitin line with the observations from our country analyses.
In models 3 and 4 we separately assess the effects of ideological distance and the interaction between party size and office experience. Ideological distance has the expected effectthe farther away from the government, the more negative an opposition party's sentiment isbut the effect diminishes when other party features are included. 3 Party size and office experience have the expected interactive effect. Given no office experience, larger parties express more negative positions then smaller ones, but this effect is mitigated as office experience increases. 4 Figure 6 plots the interaction effect based on our full Model 5. It shows that the effect of party size is negative for parties with up to 8 years of office experience in the last 30 years, which accounts for 76% of the parties in our data (19 of 25). The effect of office experience is positive for parties obtaining more than 11.2% of all seats in the chamber, which accounts for 28% of the parties (7 of 25). Therefore, for most parties, larger size leads to more negativity, and for most parties, office experience does not significantly affect position. Only the largest parties are positively affected in their position-taking by their government experience, and only the most experienced opposition parties are unaffected by size. 5

Discussion and conclusion
Our analysis shows that opposition party expressed sentiment in parliamentary debates on COVID-19 started out relatively positive towards the government's actions and policies and increasingly became more negative. Many opposition politicians cited the need for cooperation across the aisle when countries were faced with an exceptional public health crisis. Our findings are in line with those observed in military crises and possibly also the longer-term dynamics in economic crises. In the short term, one can observe a 'rally effect', while in the longer term the negative consequences of the crisis result in more negative opposition party expressed sentiment. The finding of more positive opposition expressed sentiment early on in the crisis is also in line with a 'rally around the flag' effect in public opinion that was observed, to a varying extent, in many countries (Kritzinger et al. 2021;YouGov 2020). Contrary to our expectations, we do not observe a curvilinear pattern of opposition party expressed sentimentfirst an increase, then a decrease in positive sentimentbut rather a monotonous decline. One potential explanation is that public health crisis management was not a politicised issue before the start of the pandemic and therefore did not lead to adversarial debate. Government and opposition parties mostly followed expert advice. When it became clear the COVID-19 crisis would not be over in a matter of weeks or even months, opinions on policy actions started to diverge more. Moreover, the crisis broadened from a public health situation to a social and economic crisis, which are traditionally more politicised issues. Another explanation might be that governments reached out to the opposition early on during the crisis, which resulted in positive expressed sentiment from opposition politicians. Interestingly, our analysis of parliamentary debates shows that this was particularly the case in the United Kingdom as well as in Germany; in the latter case the need for federal coordination is an important factor. In contrast, in the Dutch case we see few opposition spokespersons refer to opposition inclusion in COVID-19 policy making, at least not beyond what is considered normal. In Israel, the situation was complicated by the fact that government formation took place at that time, which somewhat blurred the lines between government and opposition. In future work on government-opposition relations during a crisis, the government's attitude to cooperation with the opposition could be explored more fully.
While we observe more positive opposition expressed sentiment early on in the COVID-19 crisis, we find little indication that governmentopposition relations have changed qualitatively due to the crisis. Partyspecific factors play out as indicated by previous research (Tuttnauer 2018). All opposition parties were more positive towards the government early on in the crisis, but relative differences between parties remained. One reason might be that parliamentary actors interpret and address the crisis and its consequences within established competitive strategies. Where critical questions were asked of the government, these largely originated from parties' existing priorities: for example, social-democratic parties highlighted social and ethnic inequality, the SNP emphasised regional autonomy and the Green parties stressed the importance of Funding