Strategic interdependence and preferences for debt mutualization in the eurozone

Abstract Existing research argues that a ‘democratic constraint’ blocks the path towards fiscal integration in the eurozone: Voters in creditor countries are fundamentally opposed to debt sharing, while voters in debtor countries are unwilling to leave the euro, which constrains the ability of their politicians to negotiate a more equitable distribution of the burden of adjustment. However, this literature neglects that preferences are strategically interdependent across countries and are affected by the type of information processed by individuals. Based on two linked survey experiments in Germany and Italy, conducted at a crucial moment during the COVID-19 pandemic, we show that Germans respond to information that highlights the possibility of a break-up of the euro as a result of Italexit by increasing their support for debt mutualization. In contrast, Italians are more affected by information about the costs of remaining in the euro than of exiting it and drastically reduce their support for the euro if they are told that austerity and structural reforms are required to remain in it. Our results help to explain why German politicians relaxed their hostility to joint debt and agreed to the introduction of the EU’s pandemic recovery fund in 2020.


Introduction
International integration is becoming increasingly contentious at the domestic level. The most visible manifestations of popular opposition to continued integration were the Swiss and British referenda of 2014 and 2016, respectively, in which a majority of citizens voted to reject the restrictions that come with membership in supranational institutions. More generally, national governments now face stronger constraints when negotiating at the international level. If, in the past, governments' efforts to further international cooperation could rely on the tacit support of citizens, they have had to face an increasingly recalcitrant public in recent years (Hooghe & Marks, 2009;Beramendi & Stegmueller, 2020;de Vries et al., 2021;Walter, 2021;Z€ urn, 2018). country. Our research design exposed respondents in Italy to a hypothetical scenario in which Italy faced a financial crisis that could lead to its exit from the common currency. They were asked to weigh the costs of a disorderly exit from the euro against the costs of implementing austerity and structural reforms as a condition for remaining. German respondents were exposed to the same scenario in which Italy faces a financial crisis, and they were asked to evaluate the costs of a possible breakup of the euro for the German economy against the costs of allowing debt mutualization in order to protect the euro.
Our results suggest that the preferences of northern citizens with regard to fiscal mutualization depend on expectations about what the counterpart will do. When evaluating the costs of euro breakup and debt mutualization simultaneously, the former have a greater impact on preferences in Germany, which dovetails with similar evidence about the preferences of German interest groups . Differently from Greek citizens in 2015 (Jurado et al., 2020;Walter et al., 2018), Italians weigh the cost of austerity reforms more heavily than the costs of a disorderly exit. Ironically, it seems that members of the eurozone need to experience a serious threat of disintegration first (with a systemic country being willing to exit) before they can integrate further.
Beyond the eurozone, our results underline the importance of taking into account citizens' perceptions of interdependencies in the international political economy when explaining public preferences about integration or disintegration. The current literature is correct to emphasize that the choices of national leaders in the international arena are increasingly constrained by domestic preferences. However, we highlight that citizens' preferences are iteratively and dynamically influenced by their perceptions of what other countries will do and by the way in which these decisions are portrayed. In other words, strategic interdependencies shape public opinion. 2 In line with the new interdependence approach (Farrell & Newman 2016), these political feedback effects need to be taken into account when explaining governments' positions in international negotiations.

Fiscal integration and Europe's democratic constraint: a brief overview
When the euro was created, intergovernmental conflicts prevented the introduction of tools for risk-sharing and debt mutualization (Dyson & Featherstone, 1999;Eichengreen & Frieden, 2001;McNamara, 1998). Northern European countries feared that they would have to pay for the profligacy of southern European member states. Therefore, they intentionally excluded any mechanism for fiscal risksharing from the Maastricht Treaty.
The euro crisis, however, revealed the problems of an incomplete monetary union. When Greece lost access to financial markets, and several other member states followed suit, the European partners had to bail them out, first through adhoc measures and then through the introduction of the European Stability Mechanism (ESM). However, the fiscal architecture of the eurozone was not fundamentally changed: Countries receiving financial assistance were required to sign a Memorandum of Understanding, committing to structural reforms and austerity policies. Moreover, the mechanisms of fiscal surveillance were strengthened through the introduction of the Fiscal Compact. Although debtor countries received help to service their debts (Schelkle, 2017), the burden of adjustment fell almost exclusively onto them (Frieden & Walter, 2017;Walter et al., 2020).
There was no shortage of ideas for an alternative resolution of the eurozone crisis, but northern states were reluctant to implement them. This resistance has been described as resting on a strong moral foundation (Carstensen & Schmidt, 2018;Hien, 2019;Matthijs, 2022;Matthijs & McNamara, 2015), but the most common explanation focuses on public opinion. In the last few decades, European policymakers have increasingly been confronted with Eurosceptic publics (Hooghe & Marks, 2009. Faced with a 'constraining dissensus', political leaders increasingly had to 'worry about the electoral consequences of their European policies' (Hooghe & Marks, 2009, p. 21).
Numerous studies show that the public in northern European countries was strongly opposed to fiscal integration. For instance, a majority of Germans were opposed to European bailouts and support for Eurobonds was particularly low in countries that would have been net contributors (e.g. Kanthak & Spies, 2018;Nicoli, 2019). Beramendi and Stegmueller (2020) argue that the opposition towards fiscal integration is rooted in domestic economic interests that diverge across the eurozone. This heterogeneity in citizens' preferences makes risk-pooling among member states difficult as 'leaders weigh heavily the preferences of their national electorates when negotiating policy' (Beramendi & Stegmueller, 2020, p. 640). Based on interviews with policymakers in Austria, Germany, and the Netherlands, Walter et al. (2020, pp. 225-229) confirm that elites were aware of the popular opposition towards fiscal transfers and that this opposition constituted the main obstacle for further fiscal integration. According to Schneider and Slantchev (2018, p. 3), Germany's strategy was 'firmly rooted in domestic politics' and 'Merkel's tough talk on Greece brought her political gains domestically'. Taken together, opposition towards fiscal integration by northern governments was 'aligned fully with an extremely hostile public opinion towards Eurobonds' (Howarth & Schild, 2021, p. 219, also see Bernhard & Leblang, 2016;B€ orzel & Risse, 2018;Degner & Leuffen, 2020).
Simultaneously, southern governments faced their own version of the 'democratic constraint'. The prolonged eurozone crisis substantially increased dissatisfaction with the EU (de Vries, 2018;Guiso et al., 2016;Hobolt & de Vries, 2016). Remarkably, however, support for the euro remained high (Hobolt & Wratil, 2015;Roth et al., 2016;Walter et al., 2018). Even in the crisis-ridden south, a majority of citizens still supported the euro despite austerity and a prolonged recession. There was a broad consensus in debtor countries that exit from the euro should be avoided at all costs, not just in Greece but also in other countries (Walter et al., 2020, Chapter 4). Many people in southern Europe thus had conflicting preferences because they supported the euro but opposed austerity policies (Clements et al., 2014;Fern andez-Albertos & Kuo, 2016;Franchino & Segatti, 2019). As citizens compare the status quo against possible alternatives (de Vries, 2018), the prospect of leaving the euro was even less attractive than austerity for citizens in the south.
In the negotiations with creditors, this popular support for the euro influenced the southern governments, which is illustrated by the events around the third Greek bailout in 2015. Jurado et al. (2020) show that while the far-left government led by Syriza was trying to renegotiate the terms of the agreement with the creditors (the 'Troika'), support for the euro remained high in Greece (around 75 percent), despite the negative consequences of austerity for the Greek population (Xezonakis & Hartmann, 2020). In summer 2015, when a majority of voters rejected the third bailout package in a popular referendum, more than threequarters of respondents wanted to keep the euro, and only 13 percent preferred exit (Walter et al., 2018, p. 982). Arguably, popular support for the euro deprived the Greek government of a credible exit option and reduced its bargaining power (Varoufakis, 2017, p. 478).
Public preferences, therefore, favored the negotiation position of creditor countries: while southern countries had no credible alternative, northern countries had no incentive to allow for greater fiscal integration. This is why Beramendi and Stegmueller (2020, p. 642) argue that 'Europe's perpetual stasis has clear and traceable democratic origins', while Schneider and Slantchev (2018, p. 28) contend that 'strong domestic opinions can lead to suboptimal foreign policies'. The strategic imbalance led to constant muddling through (Jones et al., 2016), as EU leaders took 'the path of least political resistance, keeping the euro afloat with regulatory measures while avoiding populist pressures that would arise in major treaty reform' (Hooghe & Marks, 2018, p. 117). Meanwhile, the fundamental problems of the eurozone remained unsolved (Copelovitch et al., 2016;Jones et al., 2016;Matthijs & Blyth, 2015).
In 2020, however, the debate about fiscal risk-sharing returned with a vengeance. The COVID-19 pandemic affected all European countries, but southern countries like Italy and Spain were initially hit more strongly than northern countries. Furthermore, the fiscal capacity of southern countries to respond to the economic shock was lower due to high government debt. Consequently, in March 2020, nine heads of states from the eurozone signed a letter to European Council President, Charles Michel, demanding common European bonds to finance the crisis response. Following the script from the euro crisis, northern governments, including Germany and the so-called 'Frugal Four' (Austria, Denmark, Netherlands, Sweden), initially opposed this idea. However, in a surprising turn of events, the German government shifted its position in May when Angela Merkel and French President Emmanuel Macron proposed a pandemic recovery fund, including grants to member states financed through joint European debt. The initiative paved the way for 'Next Generation EU', a e750 bn instrument to respond to the pandemic. The centerpiece of this pandemic recovery fund is the Recovery and Resilience Facility which distributes funds to member states raised through joint borrowing. Although there is no mutualization of legacy debt and transfers to the south remain limited compared to the deficits incurred due to the pandemic, this was a surprising shift in the German and European position with regard to fiscal risk-sharing.

The role of strategic interdependence in the formation of preferences
The argument about a democratic constraint neglects that choices in 'debtor' and 'creditor' countries are strategically interdependent. When a eurozone member experiences financial difficulties, creditor countries need to evaluate the costs of stabilizing a country against the risk that the country could exit from the eurozone, especially if it is economically and politically important to them (Schneider & Tobin, 2020). Debtor countries, in turn, need to evaluate the costs of staying in the eurozone against the costs of a possible exit. The former is contingent on whether creditor countries agree to 'share the burden'. In short, support for debt sharing in creditor countries depends on whether debtor countries threaten to exit, while support for exit in debtor countries depends on whether creditor countries are willing to agree to debt mutualization.
As argued above, a credible threat of rupture of the eurozone was absent during the euro crisis, due to popular attachment to the euro in the southern countries directly affected by the crisis. As implied by the literature on the 'democratic constraint', in these circumstances the unconditional preferences of the northern electorates constrain northern policymakers to oppose any move towards fiscal risk-sharing and debt mutualization. However, the COVID-19 pandemic introduced an important change in background conditions.
The epicenter of a possible new financial crisis moved to Italy, a systemically important country, which would be more difficult to 'ring-fence' than a crisis in Greece or other countries. Furthermore, already before the pandemic, Italy had high levels of Euroscepticism as a result of 25 years of economic stagnation, and the health emergency increased it further, as we illustrate later in the paper. Against this background, analyses that focus on unconditional preferences (e.g. 'are you for or against debt mutualization?', 'are you for or against remaining in the euro?') are likely to suffer from omitted variable bias. Rather, we need to study the strategic preferences of citizens, which depend on perceptions of what other countries might do when faced with a potentially highly disruptive crisis. 3 Debt mutualization, breakup, or a continuation of the status quo are all possible outcomes of a fiscal crisis, but the impact of each of these outcomes for individual countries is difficult to determine in advance. If citizens had complete information and full capacity to process it, they would objectively estimate both the costs and the probability of different outcomes under alternative scenarios, and form their preferences accordingly (Gabel, 1998). However, this is not realistic. Most people do not have clearly specified views on complex issues (Zaller, 1992). They have a number of different partially inconsistent ideas and considerations in their head, but their specific preferences, as expressed in surveys, depend on the type of information people process when forming their 'causal beliefs' (Rho & Tomz, 2017).
In this paper, we use frames to manipulate the information that individuals process. Specifically, we focus on frames that provide information about the costs associated with a country exiting or remaining in the eurozone for both debtor and creditor countries. A large literature on issue framing shows that highlighting the positive or negative consequences of a policy choice in survey experiments substantially affects individual-level preferences (Amsalem & Zoizner, 2022;Chong & Druckman, 2007;Lupia, 1994;Slothuus & de Vreese, 2010). Here, we do not use frames to selectively emphasize different features of the same policy choice, but to convey information about the costs of different options, which individuals may not consider if it was not made available to them (Rho & Tomz, 2017;Tversky & Kahneman, 1973).
As argued above, Europe's crisis resolution strategy adopted during the eurozone crisis prescribes that countries unable to roll over their debts receive financial support from the ESM in exchange for strict conditionality, including austerity and structural reforms. The unpopularity of these measures led to a decline in support for the euro (Baccaro et al., 2021; Fern andez-Albertos & Kuo, 2016; Franchino & Segatti, 2019). We thus expect that respondents in debtor countries are more likely to support euro exit if they receive information that links continued membership in the euro to the implementation of austerity measures. Conversely, we expect respondents in creditor countries to be less supportive of the continued membership of debtor countries in the euro if they receive information linking continued membership with the need for debt mutualization or fiscal guarantees, which are unpopular in creditor countries.
Hypothesis 1: Support for a debtor country remaining in the eurozone (exiting from it) is lower (higher) when respondents receive information highlighting the costs of the debtor country remaining in the euro (the implementation of austerity policies for the debtor country; debt mutualization for the creditor country).
Following a similar logic, respondents receiving information about the costs of a debtor country exiting the euro should become more favorable to keeping the eurozone intact. Exit would carry substantial costs for both debtors and creditors due to externalities and knock-on effects. These costs are difficult to specify a priori because there is no precedent of a country exiting the eurozone that could offer guidance. Nonetheless, we can speculate about the costs of exit of a systemically important country based on the existing economic literature (Eichengreen, 2010;Manasse, 2019). In the country that exits, at a minimum, there would be the costs and uncertainty associated with a currency changeover. In all likelihood, the exchange rate would depreciate, and inflation would increase, which would erode the purchasing power of non-indexed income. A severe recession and the bankruptcy of many businesses are also possible, especially if their assets are redenominated while their liabilities remain in euro. Furthermore, domestic banks' holdings of local sovereign bonds would likely lead to capital losses and bankruptcies in case of exit. At the same time, an exit would also impose severe adjustment costs on other member states, and even threaten the euro in its current form. For creditor countries, it would likely result in appreciation and competitive losses for their export industry and endanger the stability of Europe's financial system. We thus expect the following: Hypothesis 2: Support for a debtor country remaining in the eurozone (exiting from it) is higher (lower) when respondents receive information highlighting the costs of the debtor country exiting from the euro (disorderly exit for the debtor country; threat to export-led growth and financial stability for the creditor country).
The question of which types of costs, exit or stay, have the greatest impact on respondents' preferences is largely an empirical question. Nonetheless, prospect theory provides some guidance (Quattrone & Tversky, 1988;Tversky & Kahneman, 1974, 1992. It suggests that the assessment of costs and benefits is not symmetrical but depends on the perspective from which they are evaluated. Specifically, if the evaluation is conducted from the 'domain of gains', decision-makers exhibit riskaversion, which means that when given a choice between a certain and a risky prospect with the same expected outcome, they choose the certain outcome. Vice versa, if the evaluation is conducted from 'the domain of losses', decision-makers are risk-seeking and prefer the uncertain to the certain prospect even though they have the same expected outcome. 4 When applied to the hypothetical scenario of our experiment, prospect theory thus suggests that the impact of information about the costs and benefits of the eurozone will depend on whether the eurozone is perceived from the 'domain of losses' or from the 'domain of gains'. In the former case, information about its costs will weigh more heavily on preferences; in the latter case, the opposite applies. To put it differently, respondents who see the eurozone as fundamentally a good thing, worth preserving, will place more weight on information about the costs of breaking up the euro, while respondents who see the eurozone negatively as bringing losses will give more weight to information about the costs of remaining in the euro. To the extent that respondents in debtor countries have a more negative perception of the euro than those in creditor countries, information about the costs of remaining in the euro will move preferences more than information about the costs of a euro breakup in debtor countries, and vice versa for creditor countries.
Hypothesis 3a: For respondents in debtor countries, information emphasizing the costs of a euro breakup has a smaller impact on preferences than information emphasizing the costs of remaining in the euro.
Hypothesis 3b: For respondents in creditor countries, information emphasizing the costs of a euro breakup has a larger impact on preferences than information emphasizing the costs of maintaining the eurozone intact.

Data and methods
We study public preferences for reform of the eurozone at a crucial time during the COVID-19 pandemic, and we focus on one key debtor country, Italy, and one key creditor country, Germany. Italy's pre-existing high level of public debt (135 percent of GDP in 2019), the increase in the public deficit caused by COVID-19 (9.5 percent of GDP in 2020), and 25 years of economic stagnation in which Italy's growth rate has usually been lower than its interest rate, make Italy especially vulnerable to a fiscal crisis. At the same time, public support for the euro has been lower in Italy than in most other European countries (European Commission, 2019). Survey evidence, discussed later in the paper, suggests that COVID-19 caused an exogenous drop in support for the EU at the time we launched our study.
The possibility of Italexit implies a threat to the euro in its current form. As the third biggest economy of the eurozone, Italy's exit would have severe consequences for other member states. This is especially the case for Germany, which headed the creditor front during the eurozone crisis and strongly influenced Europe's crisis resolution strategy. Germany has been one of the primary beneficiaries of the eurozone because its export-led growth model arguably benefits from the single market and an undervalued real exchange rate (Scharpf, 2018;Johnston & Regan, 2016). Moreover, German public opposition to debt mutualization was an important determinant of the German government's position during the eurozone crisis, as detailed above. Yet, Germans have also been highly supportive of the euro (European Commission, 2019). We thus need to know whether they would support debt mutualization if this was necessary to avoid Italexit and a breakup of the eurozone.
We use a linked survey experiment in both countries. Fieldwork began on March 31, 2020immediately after European heads of state asked the Eurogroup (made up of eurozone finance ministers) to present a proposal for Europe's joint fiscal response to the pandemicand ended before the Eurogroup meeting on April 7 where this proposal was discussed. To allow for the completion of the surveys in a short period with representative samples, we conducted them separately with SWG in Italy and respondi in Germany. Both surveys were based on a common questionnaire, and we closely coordinated and monitored implementation. In Italy and Germany, we sampled 2,118 and 2,246 respondents, respectively, from a large pool of individuals, who were recruited online and by telephone to ensure a balanced composition of the population. 5 We used sample quotas to ensure a representative sample based on age, gender, and education. In the analysis, we further use weights to correct for deviations in our sample from the true population. The weights account for the inclusion probability with respect to region, age, gender, education, and past vote choice, but results do not depend on them (or the type of weights).
In both Italy and Germany, the survey included a factorial survey experiment. 6 In Italy, we had conducted a first wave of our survey about preferences towards the euro in October 2019 (Baccaro et al., 2021). This allows us to compare baseline attitudes towards remaining or exiting from the euro before and at the beginning of the pandemic. We made the second wave as comparable as possible to the first wave. In Italy, we asked all respondents to imagine a basic scenario in which Italy is faced with a financial crisis: Italy faces a crisis of confidence in financial markets. Capital flies out of the country; customers try to withdraw their deposits from banks; and the interest rate spread with Germany increases. As a result, the Italian government is unable to meet its financial obligations. Other European countries offer Italy a bailout package.
Before deciding whether to accept or not the bailout package, the government calls a referendum. The referendum asks citizens whether they want to stay in the euro and thus accept the bailout package, or whether they want to reject the bailout package and therefore exit the euro.
We adapted this scenario to the German case by asking respondents to imagine the following basic scenario: Italy faces a crisis of confidence in financial markets. Capital flies out of the country; customers try to withdraw their deposits from banks; and the interest rate that the Italian government has to pay to issue government debt increases. As a result, the Italian government is unable to meet its financial obligations.
The Italian government is unwilling to sign a bailout plan similar to the Greek one after the financial crisis, which would condition the disbursement of funds on the implementation of austerity measures, and is contemplating exit from the euro.
Due to its weight in the negotiations with other eurozone countries, the German government can prevent Italy from exiting the euro or facilitate Italy's exit.
In both countries, the basic scenario emphasized the strategic interdependence of countries. In Italy, the choice to exit or remain in the euro was set against the background of a European bailout plan. Even more explicitly, in Germany, the basic scenario centered on Italy not being willing to accept a bailout plan. The final sentence about the German government being able to prevent or facilitate Italy's exit acknowledged the importance of Germany in shaping the European negotiations.
After the basic scenario, we asked Italian respondents how they would vote in a hypothetical referendum about euro membership. This was a heuristic device aimed at eliciting preferences about a real decision-making situation as opposed to simply expressing an opinion (Landa & Meirowitz, 2009, p. 494). In reality, it seems unlikely (but not impossible) that a decision about Italexit would be preceded by a popular referendum. Nonetheless, the possibility of such a referendum has been repeatedly discussed by key political actors in Italy and is salient in the Italian public sphere. 7 In designing the basic scenario, we drew inspiration from the Greek crisis of June 2015, which saw the freezing of liquidity by the European Central Bank (ECB), capital flight and depositors' run on banks, a rapid increase of risk premia on government bonds, and financing problems for the treasury due to mounting interest rates. This kind of scenario is considered realistic by Italian economists (e.g. Manasse, 2019). However, our basic scenario diverts from the Greek events in one crucial respect: In Greece, the consequence of a 'no' vote was ambiguous because it was not clear whether it implied renegotiation of the bailout package or euro exit (Walter et al., 2018). We eliminated the ambiguity and created a stark choice between accepting the bailout package and remaining in the euro or rejecting it and exiting the euro. In the German scenario, the reference to Italy being unwilling to follow the established crisis-resolution script was intended to signal to respondents that acceptance of a bailout plan similar to those that had been imposed on debtor countries during the eurozone crisis was ruled out.
We randomly combined the basic scenario with two frames on the costs of 'Italremain' or 'Italexit' for Germany and Italy, respectively. 8 All frames were pure issue frames, providing no information about endorsements by parties or other actors. For Italy, the costs of Italremain scenario highlighted that the European bailout plan was contingent on the implementation of austerity policies and structural reforms. These were the policies that crisis countries had had to implement in return for European financial support during the euro crisis. In contrast, the costs of Italexit scenario highlighted the costs of a disorderly exit from the euro for Italy, i.e. inflation, reduced purchasing power, bank insolvencies, and possible retaliation by other countries. For Germany, the costs of Italremain scenario emphasized the need for Germany to agree to some form of debt mutualization to keep Italy in the eurozone. The costs of Italexit scenario concentrated on the negative implications of a collapse of the euro for the competitiveness of the German export industry.
In both countries, the combination of frames resulted in a 2 Â 2 factorial design with four different scenarios, as summarized in Table 1. 9 Respondents were shown one out of four different scenarios (three treatment groups and a control group). Afterward, we asked Italian respondents 'How would you vote in this referendum?'. In Germany, we asked, 'In your view, what should the German government do in response to this crisis?'. We use answers to these questions as our key dependent variables. 10 Support for fiscal integration vs. breakup in the shadow of the COVID-crisis We present our results in three steps. First, we examine how baseline support changed in Italy after the outbreak of the coronavirus. This allows us to validate our hunch that COVID-19 was an exogenous shock for the attitudes of Italians, increasing support for exit. Second, we analyze the results of the framing experiment. We estimate multinomial probit models and calculate average treatment effects (ATE) and predicted probabilities to test whether the frames have the expected effects. 11 Finally, we discuss some robustness checks.
Support for the euro in Italy and Germany Figure 1 illustrates that support for euro membership in Italy was considerably lower in April 2020 than in October 2019. Support for Italremain decreased from 51 percent to 34 percent, while support for Italexit increased from 30 percent to 37 percent. This means that an absolute majority for remaining in the euro before the breakout of Covid-19 turned into a relative majority for Italexit after it. Yet, uncertainty also increased: in October 2019, 18 percent indicated that they did not know how they would vote in a hypothetical referendum, increasing to 28 percent six months later. Regression results show that the changes in preferences in Italy between 2019 and 2020 are statistically significant (Table C.1 in the appendix). In Germany, there was also a close horse race in April 2020: a relative majority of 42 percent of respondents wanted Italy to remain in the euro, whereas 38 percent of respondents favored Italexit. Nearly one-fifth of the respondents were uncertain. Our surveys also show that German respondents assess the benefits of the euro more positively than Italians, and that Italian respondents assessed the euro more positively in fall 2019 than in spring 2020. In Italy, the average value on a scale from 0 to 10 (where higher values indicate higher perceptions of having benefited from the euro) was 3.78 in 2019, decreasing to 3.47 in 2020 (see Appendix C.2). In Germany, the average value was 5. 17  who reported not having benefited at all from the euro increased from 25.43 percent to 31.16 percent, while only 12.62 percent of German respondents fall in this category. Overall, these descriptive findings suggest that the pandemic was, indeed, an exogenous shock for attitudes toward the euro in Italy and that Italians are more likely to evaluate euro membership negatively, whereas Germans are more likely to evaluate it positively.

Evidence from the Italian survey experiment
In the next step, we examine the effects of the framing experiment in Italy. Figure 2 shows the average treatment effects of the two frames and their combination. The costs of Italremain frame has a strong effect on preferences, which is in line with our Hypothesis 1: informing respondents that Italy's continued membership in the eurozone is contingent on the implementation of austerity measures increases support for Italexit by 15.57 percentage points and reduces support for remain by 13.35 percentage points. Contrary to Hypothesis 2, we do not find a statistically significant effect of the costs of Italexit frame. This scenario hardly shifts preferences. Although surprising, this result is in line with research on Brexit, which finds that British voters also discounted the costs of exit in the 2016 referendum (Carreras, 2019;Grynberg et al., 2020). The results from the combined treatment are in line with Hypothesis 3a: when both frames are presented together, the costs of exit frame clearly prevails. It increases support for Italexit by 14.83 percent, which is only marginally smaller than the effect of the costs of exit frame alone.
To test whether the frames can alter the democratic majorities for or against Italexit, we also calculate the predicted probabilities for voting in the referendum by treatment (Figure 3). In nearly all scenarios, we find a relative majority in favor of Italexit. Only under the costs of exit frame (scenario 2), we find a relative majority in favor of Italremain, but the difference between exit and remain is statistically insignificant. In contrast, the frame making austerity a requirement for Italy's continued membership in the euro has a strong effect, leading to an absolute majority in favor of Italexit (52.92 percent, scenario 3). When both frames are combined (scenario 4), there is still a majority (52.09 percent) for Italexit, which indicates that the costs of austerity and conditionality prevail over the costs of exit. Overall, these results suggest that Italians would prefer exiting the euro if remaining in the euro after a fiscal crisis required austerity.

Evidence from the German survey experiment
Results from the German survey are also in line with our expectations (Figure 4). The frame highlighting the costs of Italremain for Germany increases support for exit by 9.02 percentage points, while it decreases support for remain by 10.37 percentage points (Hypothesis 1). Thus, Germans react negatively to debt mutualization. However, the frame highlighting the costs of Italexit for the German economy has a larger effect: it increases support for remain by 14.85 percentage points and decreases support for exit by 12.96 percentage points (Hypothesis 2). Importantly, when the costs of Italremain and Italexit are mentioned simultaneously, the costs of Italexit prevail and the treatment effects are very similar to the scenario where only the costs of Italexit are mentioned: support for Italy remaining in the eurozone increases by 8.61 percentage points, while support for Italy exiting from the eurozone decreases by 9.67 percentage points. This is in line with  Hypothesis 3 b and suggests that Germans are more worried about the costs of Italexit for the German economy than about the costs of debt mutualization. Figure 5 again shows the predicted probabilities associated with the different frames. As shown above, in the control group, a relative majority of Germans favor Italremain. The difference between the shares of respondents preferring remain and    exit is statistically insignificant, though. The costs of Italexit frame substantially bolsters the electoral majority for Italremain, leading to an absolute majority (57.27 percent) in favor of remain. The frame emphasizing the costs of Italy remaining in the euro for Germany leads to a relative majority in favor of Italexit (47.34 percent). When the costs of remain and exit are presented together, an absolute majority (51.04 percent) is in favor of Italremain and debt mutualization.

Robustness checks
To probe our results, we ran additional robustness checks. First, we explored heterogeneous framing effects. The implications of euro membership and debt mutualization are cognitively demanding. Education, economic knowledge, or interest in economics may condition how individuals react to the frames due to their complexity (Chong & Druckman, 2007). However, heterogeneous effects are absent for most frames and individuals react to the treatments in similar ways (Figures D.7 to D.12). Given the complexity of our frames, we also tested whether treatment effects differed by response time. The results are neither affected by respondents' response time to the entire survey nor the time that they took to read the vignettes (Figures D.5 and D.6). 12 We also tested to what extent framing effects could be contingent on national identity or attitudes towards the EU (Figures D.13 to D.16). Again, heterogeneous effects are largely absent. 13 Second, we also re-ran the analyses after dropping all respondents who indicated that they did not vote in the last parliamentary election, given that policymakers may care more about the preferences of voters than non-voters (Appendix D.5). This does not affect our results. Third, as a further test that public preferences respond to the strategic interdependence of countries, we used a different dependent variable. In Italy, we asked respondents whether they prefer to remain or exit from the euro if Germany and other European governments do not agree to debt mutualization. In Germany, we asked respondents whether they prefer a scenario in which Germany and other European governments do not agree to share debts and Italy exits the euro or one in which Germany and other European governments agree to share debts and Italy remains in the eurozone. 14 The results with this variable are similar to the main analysis (see Appendix E): Italian respondents are willing to exit the euro in the absence of debt mutualization. In turn, German respondents are willing to accept debt mutualization if the alternative is a breakup of the eurozone. This confirms that preferences depend on the expected choices of the counterpart.

The debate on coronabonds in Italy and Germany
The experimental evidence suggests that Germans become more favorable to debt mutualization when they are aware that there is a threat of breakup of the euro, relaxing the 'democratic constraint' on more fiscal integration. In this section, we document that this experimental evidence resonates with the political debate on Coronabonds in Italy and Germany in 2020. Based on a review of national newspapers, we show that German policymakers started to respond to the pandemic by following the script of the euro crisis, suggesting the intervention of the ESM. 15 However, along with concerns about the single market, they quickly began to worry about the integrity of the eurozone, especially with regard to a possible Italexit. This led them to embrace a mutualization of COVID-related sovereign debt, which corroborates the results of the survey experiment, suggesting that citizens' perceptions of strategic interdependencies are important to consider when evaluating the extent to which politicians are constrained by public opinion.
Italy's threat to 'go alone' Italy was hit by the pandemic earlier and more violently than other European countries. After other European countries blocked the exports or the transit of medical masks and other protective gear, requisitioning them for domestic use, 16 there were outpourings of indignation in the Italian daily press 17 and the emergence of a widespread perception that Italy had been abandoned by its European partners. 18 A survey fielded at the end of April 2020 found that 79 percent of Italians believed that EU support for Italy had been too little or not at all adequate during the crisis. 19 Another survey suggested that 67 percent believed being part of the union was a disadvantage for Italy, up from 47 percent in November 2018. 20 In this period, the Italian governmenta coalition between the M5S and the PDtried to convince the EU to introduce 'corona recovery bonds'. The government never explicitly threatened exit from the euro but faced with a stalemate in the negotiations in early April, prime minister Conte declared repeatedly that if Europe failed to live up to expectations, Italy would 'go alone' ('far a da sola'). 21 Although this was an ambiguous declaration, it could be read as a veiled threat that Italy was prepared to leave the euro if circumstances forced it to do so. For example, it was interpreted as such by Alesina and Giavazzi, two famous Italian economists, in a commentary in Italy's most circulated newspaper: That the Premier should address European meetings with phrases such as 'Ready to go alone' is not only counterproductive but absolutely lacking in credibility. How can Italy threaten to leave Europe and the euro? 22 Similarly, the Financial Times (FT) worried about the risk of Europe 'losing Italy', writing that 'there is a rising feeling among even its pro-European elite that the country is being abandoned by its neighbors … Many in Rome now feel that unless bold action is taken by northern European countries, they risk Italy turning its back on the European project forever'. 23

German reactions to the threat of Italexit
In Germany, politicians initially rejected calls to introduce Eurobonds. For example, Markus S€ oder, the leader of the Bavarian CSU, gave a 'clear no to Eurobonds'. 24 This echoed Merkel's statement from the eurozone crisis, when she had said that there would be 'no Eurobonds, as long as I live'. 25 Instead, the German government argued that countries like Italy and Spain should avail themselves of the existing institutional channels, i.e. apply for an ESM loan. 26 Rapidly, however, concerns about the integrity of the eurozone emerged. Stories about the increasing unpopularity of the EU in Italy featured prominently in German media. German newspapers cited results from the Italian opinion polls mentioned above, showing a clear drop in approval for the EU 27 and a massive fall in Germany's reputation in Italy. 28 Several news outlets argued that this rise in Euroscepticism could endanger the euro and speculated about the consequences for Germany. 29 For example, the Handelsblatt wrote that 'Italy and Spain could soon lose the confidence of the financial markets and become the target of speculative attacksand this would also put the euro itself at risk … A return to the deutschmark would be a disaster for the German export industry'. 30 Consequently, beginning in April, German policymakers became increasingly worried about the EU coming apart (Schramm, 2021, p. 10). CDU executive board member Elmar Brok declared that if Germany refuses Eurobonds, 'the Le Pens and Salvinis will scream for joy', emphasizing that there would be no post-crisis recovery in Germany if major EU countries fall to radicals. 31 Other policymakers began to argue that 'solidarity' with other European countries was in the economic selfinterest of Germany. For example, foreign minister Heiko Maas said that 'as an export-oriented economy, solidarity is also in our own interest: if large parts of the EU are deep in crisiswho will buy our products?'. 32 Similarly, finance minister Olaf Scholz highlighted Germany's position as an Exportnation 33 and the close integration of Europe's economies, 34 while Merkel argued that 'we will only do well in the long run if Europe does well'. 35 They were influenced by German businesses, which worried about the lack of demand for exports and their supply chains, which had been disrupted during the first lockdown. Moreover, the shift in Germany's position gained further traction after the German Constitutional Court declared parts of the ECB's bond-buying programs unconstitutional in early May. Concerns arose that the agreed volume of the ECB's pandemic emergency purchase program (PEPP) could prove insufficient but that an expansion of it would be politically contentious given the court's decision. 36 Therefore, only two weeks after the court's ruling, Germany abandoned its opposition to debt mutualization. Together with Macron, Merkel proposed a pandemic recovery fund that would be financed by joint European debt and would include transfers to EU member states that were hit particularly hard by the pandemic. With this initiative, the German government aimed to repair broken trust, particularly among the Italian public. 37 One senior advisor quoted in the FT said that 'the big concern is that the economic crisis will destroy the European single market and even threaten the future of the EU'. 38 Reactions to the Merkel-Macron Plan in German media and politics were relatively favorable. 39 They generally regarded the plan as necessary to avoid the political and economic collapse of the euro and the EU. Eckhardt Rehberg, a leading CDU parliamentarian, argued in parliament: 'Doing nothing would be disastrous not only for us as the Federal Republic of Germany À 60 percent of our exports go to the EUbut it would also be disastrous for Europe'. 40 Even Friedrich Merz, a conservative politician from the right wing of the CDU, argued that 'help to our partners is also in our own interest'. 41 The decision to introduce the recovery fund was also directly linked to public opinion in Italy. S€ oder, who had initially refused debt mutualization, argued in May that 'a disintegration of Europe and thus of the single market would be the far greater risk in financial terms alone … If we hesitate now or underestimate the psychological impact [of the pandemic], we risk a second Brexit in Italy'. 42 Similarly, Armin Laschet, who was later elected as CDU leader, said that 'Eurosceptic sentiments have increased among many people in Italy. We have many centrifugal forces … We must prevent the north and south from drifting apart'. 43 From a critical perspective, another leading CDU politician was quoted as saying that 'one gets the impression that the aim here is to counteract Germany's waning popularity [in other European countries] with money'. 44 After intense European negotiation, the plan resulted in the decision of the European Council in July 2020 to set up NextGenEU. Scholz called this Europe's 'Hamiltonian moment'. 45 He explicitly interpreted it as an entry into a fiscal union and stated that 'we will continue along this path'. 46 Although this may be an overstatement, German policymakers reframed the debate by focusing on the risks of a possible euro breakup, avoiding a popular backlash against the first concrete step towards debt mutualization.

The role of COVID-related solidarity
The above reconstruction shows that the risk of a euro breakup featured prominently in the German debate. However, at the time our survey was fielded, there was also a surge of COVID-related support for Italy in the German public discourse. This raises the possibility that altruistic, rather than cost-benefit considerations, explain Germany's support for common debt (Kuhn et al., 2018;Kuhn & Kamm, 2019). While COVID-related solidarity does not necessarily affect the treatment effect of our informational frames, it may inflate the size of the majority in favor of debt mutualization.
We are unable to rule out this possibility. However, we are able to measure the impact of priming respondents to think about the effects of COVID and compare this impact to the effect of highlighting the costs of Italexit for Germany. Our experiment included an additional frame that focused on the COVID-19 pandemic as the trigger of the crisis and absolved the Italian government from responsibility for fiscal deterioration, attributing it to the health crisis. Specifically, the frame preceded the vignette describing the basic scenario and went as follows: The corona crisis has forced the Italian government to significantly increase public expenditures, both to reinforce health care infrastructure and to contain the consequences of the recession. This has led to a large increase in the Italian public deficit as a share of Gross Domestic Product and a downgrade of Italian government bonds by rating agencies. As a consequence, now [the basic scenario follows].
We randomly combined this frame with the basic scenario and the two other frames highlighting the costs of Italremain and Italexit. This resulted in four additional experimental groups as shown in Table 2 for which we recruited 2,265 additional German respondents. Figure 6 plots the average treatment effects. The results show that highlighting the COVID-19 pandemic as the cause of the crisis reduces support for Italexit by 7.15 percentage points relative to the control group. However, the COVID frame does not have a statistically significant effect on the support for Italy's continued membership in the euro and the negative effect of the COVID frame on support for Italexit is smaller than the cost of Italexit frame shown in Figure 4. When both appeals are combined, the effect is only marginally larger than appealing to Germany's self-interest alone (-15.27 compared to À12.96 percentage points). In combination with other frames, the COVID frame hardly matters, except when coupled with the frame that highlights the costs of Italy remaining in the eurozone. In this case, the two frames cancel each other out: the marginal effect on support for Italexit decreases from 9.02 to 1.12 and becomes statistically insignificant. When the COVID frame is added to the combined costs of Italexit and costs of Italremain frame, it does not change the results. German preferences are thus more affected by the costs of Italexit than by sympathy with the plight of COVIDstricken Italy.
In Italy, we also included a similar COVID-frame in the experiment, testing to what extent the pandemic influences preferences towards the eurozone. The wording of the frame and its treatment effect are shown in Appendix F (see Figure F.2). The results indicate that the COVID frame has a positive effect on the likelihood that respondents support remaining in the eurozone but that it does not undermine the treatment effects of the other two frames. Moreover, in Italy, we had also included the cost of Italremain frame in our previous survey from October 2019. The frame had the same effect in this survey, and the difference in the effect size between the two waves is statistically insignificant (see Appendix F). Overall, this suggests that our experimental results are not only influenced by the unique circumstances of the COVID-19 pandemic but that they are likely to hold beyond it.

Conclusion
The COVID-19 pandemic has brought about a large increase in public debt. In low growth Italy, the level of public debt is especially worrisome, and may eventually be perceived as too high by financial markets in the absence of external support.
Other countries, such as Spain, also had to increase their debt levels. It is thus not surprising that renewed requests for debt mutualization emerged in spring 2020.
In explaining why European politicians rejected any type of cross-country sharing of fiscal risks during the euro crisis, political scientists introduced the notion of a 'democratic constraint', which attributes the lack of progress towards fiscal integration to the constraining role of domestic public opinion. Citizens in northern countries are strongly opposed to debt sharing or fiscal transfers, while citizens in southern countries are unwilling to leave the eurozone, which makes it more difficult to distribute the burden of adjustment more symmetrically between debtor and creditor countries.
However, this literature has not paid sufficient attention to the role that strategic interdependence plays in shaping preferences. As suggested by our survey experiment, for German respondents, the costs of a possible breakup of the euro loom larger than the costs of debt mutualization. When they are given an opportunity to consider both costs simultaneously, a majority in favor of fiscal risk-sharing emerges. In turn, the preferences of Italian respondents are more strongly affected by the prospect of having to implement austerity and structural reforms in order to stay in the euro than by any consideration of the negative consequences of a disorderly exit.
To be clear, we are not arguing that the previous literature was wrong to emphasize the democratic constraint. In fact, we think that it captures the strategic situation of the early 2010s rather well. In the absence of a credible threat of exit, northern politicians responded to the unconditional preferences of their electorate not to extend financial assistance to southern countries beyond what was strictly necessary to save the euro (Copelovitch et al., 2016;Degner & Leuffen, 2020;Jones et al., 2016). However, with the epicenter of the crisis moving to Italya more Eurosceptic country than Greece and a larger systemic threatthe strategic situation has changed, which affects public preferences in northern countries as well. The results of our survey experiment suggest that when a credible threat to the stability of the euro emerges, public opinion in the north becomes more favorable to risksharing and ceases to be a hard constraint for fiscal integration. This dovetails nicely with similar findings of the preferences of interest groups in Germany, which also prefer some risk-sharing to the collapse of the euro .
Increased awareness of the systemic risks for the integrity of the eurozone helps to explain Germany's surprising shift in the negotiations over Coronabonds. The position of German elites in the wake of the pandemic was different from a few years before when they had staunchly rejected all proposals for debt mutualization. Initially, the German government suggested that countries like Italy and Spain apply for loans from the ESM, but eventually, it became a strong supporter of the European pandemic recovery fund, which includes elements of joint fiscal liability that resemble Eurobonds quite closely. Newspaper evidence suggests that German politicians responded to the threat of a breakup of the euro and took the reaction of the public in southern countries, especially Italy, into account when conceiving their response. Although there was a surge in pandemic-related solidarity, our results show that the consideration of the negative consequences of a euro breakup was sufficient to generate a majority for risk-sharing in Germany.
A few remarks about the scope conditions of our argument are in order, though. For our mechanism to apply, it is necessary that the threat of euro exit is issued by a systemically important country. A sudden negative shift in public opinion in smaller countries is unlikely to have the same effect. Furthermore, for preferences in northern countries to shift towards greater support for fiscal risk-sharing, the majority has to see membership in the euro as a positive arrangement that is worth preserving; in other words, it has to assess it from the 'domain of gains'.
Moreover, this paper does not try to explain how debt mutualization comes about. Instead, it makes a more circumscribed argument that public opinion in northern countries is not necessarily an obstacle to it. Put differently, favorable public opinion in key northern countries is a necessary condition for reform of the economic architecture of the eurozone, but in all likelihood, it is not a sufficient condition. Reform still has to overcome a large set of veto points, both in individual European countries and within European institutions. Relatedly, we do not make an argument about a long-term shift in unconditional German attitudes toward risksharing, but more narrowly, that German public opinion is more sensitive to the framing of strategic interdependence than previous research has allowed for.
Finally, we do not want to imply that the recovery fund or even full-fledged Eurobonds would solve all problems of fiscal sustainability in the eurozone. In fact, the pandemic recovery fund may be too small to provide a meaningful economic stimulus. What really matters for the solvency of high-debt countries is a supportive monetary policy with the central bank that commits to putting a floor on bond prices when financial markets lose faith in them (De Grauwe, 2013). The ECB has clearly moved in this direction during the pandemic by launching the PEPP. However, the European treaties explicitly prohibit the ECB to act as an 'overdraft facility' for member states and the legitimacy of the ECB's bond-buying programs is regularly contested by the German Constitutional Court. Although it is probably not sufficient to stave off a possible fiscal crisis, a joint fiscal capacity is thus an important step towards putting the euro on more solid ground.
Regardless of whether or not fiscal tensions in the eurozone resurface, our findings have far-reaching implications for how we study preferences for international integration (and disintegration) beyond the issue of debt mutualization in the eurozone. Although the eurozone is the most extensive form of international integration, economic interdependencies are widespread (Farrell & Newman 2016), and global challenges such as climate change or rising geo-economic tensions over energy and raw materials are likely to exacerbate the need for international negotiations to address these interdependencies. Our findings shed light on some of the difficulties associated with such negotiations.
First, further international integration is more readily accepted when citizens in countries that are 'winners' of economic integration are primed to take the costs of disintegration into account. Second, citizens in 'losing' countries tend to underestimate the destructive potential of economic disintegration and take a more riskseeking approach to disintegration. This is a significant source of instability, as some literature on the causes of Brexit suggests (Carreras, 2019;Grynberg et al., 2020). Third, in a world of partial information and radical uncertainty, the assessment of the costs of benefits of integration and disintegration depends on the type of information processed by individuals to form 'causal beliefs' (Rho & Tomz, 2017). In this context, political elites are less constrained by public opinion than they sometimes pretend to be (de Vries et al., 2021;Gabel & Scheve, 2007). Preferences respond to information and arguments, creating room for national leaders to shape the public's response.
Notes 1. In this paper, we use the terms 'debt mutualization' and 'fiscal risk-sharing' interchangeably. 2. We use the term 'public opinion' to refer to the aggregate of individual preferences on a given topic. 3. Appendix B formalizes the argument about the strategic interdependence of preferences with a simple game theoretical model. 4. Drawing on prospect theory, Carreras (2019) argues that British voters in districts facing long-term economic decline were more likely to evaluate the economic consequences of Brexit positively and more likely to vote for Brexit (the risky option). 5. For further information on our survey, see Appendix A. 6. The experiment was pre-registered and received Institutional Review Board (IRB) approval. 7. The referendum would have to take the form of a consultative referendum (referendum di indirizzo) in the Italian legal systems and its organization would have to be authorized by a constitutional law approved by two-thirds of both chambers. A consultative referendum was held on June 18, 1989, when Italians overwhelmingly approved giving the European Parliament a mandate to draw up a European constitution. In June 2020, a citizens' initiative was launched for the organization of a similar referendum, see https://www.gazzettaufficiale.it/eli/id/2020/06/13/20A03215/sg (accessed on December 7, 2020). 8. For Italy, the basic scenario and the costs of Italremain frame were essentially the same as in the October 2019 survey; the costs of Italexit frame was new. For the exact wording of these frames, see Appendix A. 9. To be as realistic as possible, our frames combine various elements. While we are able to identify overall treatment effects of the frames thanks to randomization, which ensures exogeneity by design, we are unable to specify the role that different elements of our frames play. For example, for the austerity frame in Italy, we cannot determine to what extent any shift in preferences is due to easier rules for layoffs, expenditure cuts, privatization, etc. This is acceptable, in our view, because these elements have historically been bundled together in bailout packages. 10. The Italian dependent variable has four categories: accept the bailout and remain in the euro, reject the bailout and exit the euro, would not vote, and don't know. We merge respondents from the last two categories to simplify the analysis. The German dependent variable has three categories: prevent Italexit; facilitate Italexit; don't know. 11. Robustness tests which include control variables (age, age squared, gender, education, subjective income, national identity, economic knowledge, and region) do not change the results (Tables C.3 and C.4). The results also remain the same if we use linear probability models instead of multinomial probit models (see Appendix D.2). 12. To improve the quality of our data and to increase the likelihood that respondents would engage with the survey, they had to stay on the page with the vignette for 30 seconds. An attention check, moreover, was used to screen out inattentive respondents (also see Appendix A.1). 13. Individuals with an exclusive national identity and a negative assessment of the EU are somewhat less affected by the costs of remain treatments. This finding can largely be attributed to ceiling effects, given that support for exit is already high among those individuals in the control group. 14. These questions were asked immediately after the question which we used in the analysis above.