On 11 January 2010 the lie became official: Eurostat—the agency responsible for statistical information in the European Union (EU)—published a report that questioned the figures about national debt and budget deficits that the Greeks had supplied. The subsequent drastic austerity measures the Greek government implemented provoked determined popular resistance and in less than two months the country was engulfed in often violent protests against higher taxes and deep cuts in the public sector. The leaders of the EU scrambled to stem the crisis in cooperation with the International Monetary Fund (IMF) but they could agree on only a relatively modest emergency loan. In late April the credit rating agencies downgraded Greek government bonds to junk, and the financial panic began to infect other Eurozone members. The crisis was threatening to turn into a catastrophe that could unravel the entire Eurozone, and an increasingly vocal chorus of politicians, leaders in the banking and financial industry, and economists pressed for an immediate (and very large) bailout package.
Stunningly, the lone holdout that fiddled while Rome burned was none other than Germany—the country that was the primary beneficiary of a stable Eurozone and that correspondingly stood to lose the most from its collapse. Because German banks had invested heavily in the debt the Greek government was threatening to repudiate, the German government's foot dragging was indeed puzzling. By the time Germany finally came around in early May, the crisis had deepened and spread: the overall cost of the package had ballooned to more than twice the original estimate; Germany's share alone was nearly as high as the total original amount the EU had been set to provide.
Why was international cooperation on the financial bailout so difficult to achieve even in the usually cooperative context of the EU? Why was the main obstacle to this cooperation the country that was (and still is) among the most keen on the Eurozone? Our existing explanations of international cooperation cannot answer these questions. As we document, the evidence is not consistent with theories that explain the failure to cooperate as arising from incentives to free-ride in the provision of public goods, the absence of institutions that provide information and enhance coordination or the credibility of commitments, attempts to coerce others into granting more favorable terms, or constraints imposed by more hawkish legislatures. An alternative explanation, popular in the press and among politicians at the time, centers on Merkel's fears about crucial elections that could determine whether her coalition would keep its federal dominance. It is unclear why voters would fail to see through a delaying tactic, and how a domestic conflict over the desirability of a policy affects cooperation at the international level. Somewhat astonishingly, we have no theories of how this mechanism is supposed to work in such a context, even though cooperation failures regularly happen even without serious distributional conflict.
We develop a game-theoretic model of the four-way interaction between two governments that must coordinate a response to a crisis affecting both countries but that also must face the polls domestically with an electorate that might be uncertain about what response is necessary. We analyze how the potential domestic conflict over a particular policy's desirability interacts with the desire to cooperate among the governments under asymmetric information. We show that the data are consistent with the equilibrium that can rationalize delay for electoral reasons, and that it was precisely because the Eurogroup governments were widely known to be quite supportive of the Eurozone system that they could not have acted fast enough or aggressively enough to contain the crisis and instead opted for policies that ended up endangering the very system they benefited from. Paradoxically, had Germany been less enthusiastic about the euro, Merkel could not have employed dilatory tactics and would have been able to persuade the skeptical German voters that a bailout was necessary by the end of April. The model can account for the delay, the sudden change of course, and the subsequent clobbering at the polls.
On the empirical side, we aim to adjudicate among rival explanations of the German government's puzzling behavior. We offer evidence that supports an interpretation of the German strategy that is firmly rooted in domestic politics. Aside from intellectual curiosity, there are important reasons to get the story right because the policy implications one draws from this episode differ fundamentally depending on one's interpretation. If Merkel had simply made a mistake, then there would be little to learn from this episode. If Merkel had played a traditional war of attrition to obtain better terms from Greece and the Eurozone members, then one could safely ignore domestic politics when it came to international policy. But if she delayed because of elections, then we would need to pay closer attention to the way domestic electoral issues shape international behavior.
On the theoretical side, we specify a mechanism that can explain a strategy of delaying unpopular policies until after the elections without relying on irrational voters or uncertainty over the postelection government policy. This mechanism relies on the voters’ uncertainty about the appropriateness of the policy that has already been put in place by the incumbent, and their attempts to make inferences about it. We show how strategic information transmission can occur in a multilateral setting where two governments with somewhat mixed motives for cooperation and potential for collusion have to cope with their respective electoral concerns. We also show that the presence of a second signaling actor can serve as a constraint, and that even when distributional conflicts are minimized, informational problems can translate into serious policy failures.
Our analysis has several broader implications for international relations theory more generally. The emphasis on signaling instead of distributive politics shows how the lack of transparency in international negotiations can create a “democratic deficit” that results in international cooperation that citizens don't want. However, it also reveals that the presence of a competitive domestic electorate can enable governments to implement policies that the citizens want under circumstances when that would have been impossible had they been unconstrained. In particular, preference heterogeneity among the governments makes credible signaling possible, which allows them to cooperate without suffering a domestic backlash that would otherwise have occurred. While there are clear limits to what voters can achieve with an instrument as blunt as removal from office, there are nevertheless occasions where the threat to do so can be consequential. Whether for good or bad depends on what the citizens already believe, which in turn constrains how much the government can shift their beliefs.
Domestic Politics and International Cooperation
How are we to understand situations where international cooperation clearly failed, at least for a while? If the issue is a public good, then our theories say that cooperation might fail because of incentives to free ride, high transaction costs, or inability to coordinate effectively or to commit credibly. If the issue occasions distributional conflict, then cooperation might fail because veto-wielding domestic constituencies could be unhappy with the deal their government has worked out.Footnote 1
Germany's failure to cooperate in the bailout until it was almost too late presents a puzzle to these theories. The Eurozone members worked in the dense institutional environment of the EU, had ongoing frequent interactions that involved multiple issues, faced low transaction costs, and shared information almost compulsively. As we detail in our study, whatever limited use of coercive tactics was made ended well before the crisis escalated, and there was no evidence of attempts to free ride on the efforts of others. Contributions to the bailout, like most other financial matters in the EU, were tied to the size of the economy of individual members, and there was little room to negotiate deviations from existing European Central Bank (ECB) formulas. There were certainly disagreements among the creditors—participation of the IMF, austerity measures in Greece, and loans on nonconcessionary terms—but these were resolved in principle as early as March and in practice by mid April, yet Germany still refused to cooperate for several crucial weeks. Despite parliamentary debates in the two largest contributors, Germany and France, the legislatures were not a constraint (in fact, the opposition in Germany was pushing the government to introduce the appropriate legislation).
One could focus on the citizens rather than legislators and on elections rather than ratification as the relevant constraint or motivator for governments. For example, a government could persist with a policy it knows to be bad out of fear that trying to alter it would reveal its incompetence and result in electoral defeat.Footnote 2 But since this argumentation considers only the domestic aspects of foreign policy, the analysis has no foreign government whose behavior must be taken into account: no problems of international cooperation emerge and no complications arise from that government having to be responsive to its own citizens.Footnote 3 What is needed is a mechanism that could connect international cooperation with electoral incentives in an environment where the citizens use cues from the behavior of their own government and its foreign partner to form opinions about the desirability of retaining their incumbents. But we have no such theory, and its absence is made even more glaring by recent studies that demonstrate that public opinion can constrain international cooperation during electoral periods on other issues as well, especially when the issues are salient.Footnote 4
In this way, a particular historical puzzle has identified a theoretical lacuna that we aim to fill. Upon some reflection, it is not hard to see how the need for such a mechanism can arise more generally. Consider any situation where governments must cooperate on some international policy, where the distribution of costs is basically clear (because of existing agreements or because the policy is governed by the rules of an international organization), and where incumbents face domestic elections but voters are uncertain about the desirability of that policy. Some examples of such policies are peacekeeping missions, multilateral foreign aid, environmental protection, climate change, and financial rescue packages. The government's fundamental problem is not to negotiate a better deal for itself but to persuade its own citizens either that the policy is necessary (when it wishes to implement it) or that it is not (when it wishes to avoid it), to free itself for the policy stance it prefers while simultaneously securing its reelection.
The international dimension complicates this calculus because the foreign actor, which is responding to its own domestic concerns, might act in a way that prevents the government from signaling credibly enough to move the beliefs of its voters in the direction it wants. Sometimes, the fear of electoral defeat might lock the government into a policy it did not want and, more importantly, that the voters would not have wanted either. At other times, it might keep the government disciplined enough to implement the policy the citizens prefer even though the policy is contrary to its own wishes. It could also be the case that the foreign actor's behavior unfetters the government so it could implement its desired policy and retain office even though voters would have opposed the policy had they been sufficiently informed about it.
Disentangling the conditions that give rise to these various outcomes requires one to analyze the incentives of the governments, the motivations of the citizens, and their interaction. In our model, two governments are faced with a situation whose harmful effects might require taking a (potentially cooperative) costly action to ameliorate. The citizens in each country want the action taken only if the situation warrants it but are not sure whether this is the case. They do know that their governments might have proclivities to act in circumstances they would not want them to, and they also know that the governments have better information about the necessity of taking action. The citizens, then, wish to furnish the government with appropriate electoral incentives by threatening to keep in office only an incumbent whose behavior was in line with their preferences. For such an electoral threat to work, the citizens must form some beliefs about the appropriateness of what their government has (or has not) done, and the problem is that citizens do not have much information to work with. All that we can observe is whether the government takes the action; the policy effects are not observable until after the elections. For their part, the governments wish to signal that their behavior is appropriate but since they all want to stay in office their ability to signal credibly is severely compromised. Thus, the model incorporates a cooperation problem between the governments, an agency problem between the citizens and their government, and a signaling problem between the government and its citizens.
We focus on results from the model that are directly relevant to the empirical puzzle we set out to resolve. Space constraints prevent us from exploring the rich set of insights that the full analysis reveals but we have documented them in the online supplements.Footnote 5 Generally speaking, it is quite difficult for the citizens to incentivize the governments through an instrument as blunt as elections. For wide ranges of the parameters, the interaction must involve some sort of policy failure where the governments behave contrary to the wishes of the citizens. At one extreme is the “false-positive” equilibrium where the governments act regardless of the necessity of doing so. This can happen when the citizens in both countries strongly believe that action is appropriate, which allows the governments to take advantage of the favorable circumstances and (cooperatively) implement the policy they desire (Proposition 2). If the citizens in only one of the countries hold this belief, then a second problem is added to the policy failure: not only does their own government act even when it is not supposed to, but it can be forced to bear the cost of the policy by the other government (Proposition B). In this “burden-shifting” equilibrium, the domestic agency problem gets exported as an international free-rider problem. Finally, if the citizens in both countries strongly believe that action is inappropriate, the electoral incentives become truly perverse because they can end up blocking international action even when it is necessary. In this “false-negative” equilibrium, which can exist only if the costs of inaction are not too high, governments become prisoners of citizen expectations: since they get punished even when they do the right thing, they do the wrong one (Proposition 3).
The Model
Two countries, i ∈ {1, 2}, are faced with a crisis that can potentially require costly measures to resolve. The timing of the game is as follows: the governments, G i , observe the severity of the crisis they are dealing with and simultaneously decide whether to act or not. The median voter in each country observes these public actions and the voters simultaneously decide whether to retain the incumbent. Voting is costless. After the elections, the (possibly new) governments again decide whether to implement crisis policy, after which the game ends and payoffs are realized.
Economic Environment
Without a policy to stop it, a crisis can be either mild, in which case it inflicts on country i economic damages worth θ i > 0, or serious, in which case it inflicts damages w i θ i with w i > 1. Citizens and governments are equally sensitive to economic damages. The governments know the type of crisis they are dealing with but the citizens in both countries do not: they believe that it is serious with probability s ∈ (0, 1) and mild with complementary probability. This prior is common knowledge. The results do not depend on the governments being fully informed, just that they have better information than the citizens. Whereas a mild crisis fizzles out without a government action, a serious crisis continues to inflict cumulative damages until someone acts to stop it. If at least one of the governments acts prior to the elections, then the crisis will be resolved regardless of its type. If neither acts, then the mild crisis will resolve itself after the elections but the serious crisis will deepen.
The total financial cost of a crisis policy is C > 0. Consistent with our desire to model cooperation under existing distributional rules, if the governments act together, each country pays α
i
∈ (0, 1) of the total cost, with
$\sum \alpha _i = 1$
. If G
i
acts on its own, the country bears the entire cost, α
i
= 1.Footnote
6
Whereas the citizens of country i pay costs in full, α
i
C, its government could either be as sensitive to these costs as they are or less so. Letting t
i
∈ {1, δ} denote the type of G
i
so that the government pays t
i
α
i
C when it participates in a bailout, we call a government nationalist when t
i
= 1 and internationalist when t
i
= δ ∈ (0, 1). The government's type is common knowledge.Footnote
7
When it comes to the crisis and the reaction, the different sensitivity to the financial cost of the policy is the sole source of preference divergence between the government and its citizens:Footnote 8
Assumption 1
Citizens in each country want the governments to intervene if, and only if, the crisis is severe even when there is an agreement to share the financial costs: θ i < α i C<C < w i θ i .
This assumption also implies that regardless of the government's type, both the government and its citizens prefer to have an international cost-sharing agreement in place if that government is going to implement a crisis policy. If they expect the other government to implement the policy, then they have an incentive to shift the entire burden to the other country and reap only the benefits, raising the specter of free riding.
Political Environment
Governments value being in power, which we represent by adding 1 to their payoffs if they are reelected. Citizens value that their government behaves according to their preferences. Since citizens are not informed about the nature of the crisis, they can use only the observable behavior of the governments to make inferences about the desirability of that behavior. In particular, they form posterior beliefs about the type of crisis and then ask whether their government's action was appropriate or not. They then reward or punish the incumbent depending on this inferred behavior. There are four contingencies in which citizens of the two countries can find themselves when they vote (since they have a common prior and any new information that might be revealed from the governmental actions is symmetric, the posteriors would have to be the same).Footnote
9
Let
$s_{a_1 a_2} $
be the citizens’ common belief that the crisis is severe when they observe government i taking action a
i
∈ {0, 1}. For example, s
01 denotes their belief after a unilateral action by G
2. Citizens credit the government that acts in proportion to their belief that the crisis is serious, and the government that does not act in proportion to their belief that the crisis is mild. In our example, G
2 will be credited with s
01 whereas G
1 will be credited with 1 − s
01.
When citizens apportion credit, they compare their posterior beliefs to what they expect to get from the alternative government they could select, e i ∈ (0, 1). This baseline expectation captures how contested the elections in country i are expected to be. Very low values represent cases where the incumbent is favored to win the elections whereas very high values represent cases where the incumbent is compromised and unlikely to win. Intermediate values represent competitive elections where neither has a clear advantage.
Payoffs
Payoffs are realized at the end of the game, and are as follows.
Multilateral action
The crisis is resolved regardless of type, no economic costs are incurred, the financial costs are shared, and no further action is taken after the elections. The citizens in i obtain a payoff of s 11 − α i C if they keep the incumbent and e i − α i C if they replace it. The government in country i gets 1 − t i α i C if it is reelected, and −t i α i C if not.
Unilateral action by G i
The crisis is resolved regardless of type, no economic costs are incurred, the financial costs are borne entirely by country i, and no further action is taken after the elections. The citizens in i get a payoff of
$s_{a_1 a_2} - C$
if they keep the incumbent and e
i
− C if they replace it, whereas the citizens in −i get a payoff of
$1 - s_{a_1 a_2} $
if they keep the incumbent and e
−i
if they replace it. The government in country i gets 1 − t
i
C if it is reelected, and −t
i
C if it is not. The government in country −i gets 1 if reelected, and 0 if it is not.
No action
If the crisis is mild, it is resolved, θ i economic costs are incurred, and no financial costs are incurred. The citizens in i obtain a payoff of 1 − s 00 − θ i . The government obtains 1 − θ i if reelected and −θ i if it is not.
If the crisis is serious, it deepens, and w i θ i economic costs are incurred. Since the severity is now revealed and citizens always want such crises acted upon, we assume that whatever governments are in place will reach an agreement on multilateral action, and the costs of such a program will be distributed according to the existing fixed rule. The citizens in country i get a payoff of 1 − s 00 − w i θ i − α i C. The government in country i gets a payoff of 1 − w i θ i − t i α i C if reelected and −w i θ i − t i α i C otherwise.
Preference constraints
We can now define the preferences of the governments more precisely so that elections become meaningful in the model:
Assumption 2
A nationalist government strictly prefers to cooperate in a multilateral policy if doing so ensures its reelection and if it expects to lose office after unilateral action by the other government: α i C < 1.
Assumption 3
All else equal, an internationalist government strictly prefers to intervene unilaterally in a mild crisis rather than to allow it to continue, but strictly prefers to allow it to continue if doing so ensures its reelection and if acting unilaterally results in its removal from office: δC < θ i < 1 + δC. Note that (A1) and (A2) together imply that θ i < 1 as well.
Equilibrium Refinements
The solution concept is weak perfect Bayesian equilibrium, which requires only that strategies are sequentially rational given beliefs and that beliefs are consistent with the strategies and derived by Bayes rule whenever possible. These requirements do not put any meaningful restrictions on admissible beliefs after events that are not supposed to occur when equilibrium strategies are followed, which essentially permits any subsequent behavior to be rationalized. Since expectations about actions after zero-probability events can be crucial in supporting equilibrium behavior, we would like to ensure that these beliefs are at least plausible. To this end, we shall require that the assessment satisfies something analogous to the Intuitive Criterion:Footnote 10
Definition 1
An equilibrium is intuitive if (a) there exists no deviation that can profit only the deviating player only when the crisis is of a particular type given that the citizens infer that the crisis is of that type, and (b) for any deviation that can unilaterally induce an outcome with positive probability only when the crisis is of a particular type, the citizens infer that the crisis is of that type.
Weak perfect Bayesian equilibria are merely a subset of Nash equilibria, and as such define rationality in a strictly individualist manner: the equilibrium requirements eliminate strategy profiles vulnerable to unilateral deviations. Although this definition of rationality might be appropriate when it comes to the citizens in the two countries who cannot be expected to coordinate to deviate together, it is less persuasive when it comes to the two governments. Since governments can meet in private, they could conceivably conspire to hide information from their citizens. In the model, citizens have only the actions they can observe to go on when making inferences. But what if governments collude to take advantage of this? We shall require that the equilibrium be immune to such collusion:
Definition 2
An equilibrium is collusion proof if there exists no group deviation by the governments such that (a) the payoffs from the deviation Pareto-dominate the equilibrium payoffs, and (b) no government can benefit from deviating from the collusive agreement.
Analysis
We begin by establishing our benchmark case: an equilibrium in which governments agree to a multilateral action only when the crisis is serious and do nothing if it is mild. This is the behavior citizens want, so we shall call this the citizen-preferred equilibrium (CPE). In it, the governments are always rewarded with reelection following multilateral action because the citizens believe that the action was appropriate. Unfortunately, as the following proposition shows, this happy state of affairs is unlikely unless both governments are nationalist.
Proposition 1
The citizen-preferred equilibrium can always be supported in a nationalist dyad, but can be supported in internationalist or mixed dyads only when governments are jointly vulnerable electorally (e 1 + e 2 ≥ 1). It is intuitive in all dyads but collusion-proof only in nationalist and mixed dyads.
This result establishes somewhat dim prospects for disciplining governments through electoral sanctions.Footnote 11 Internationalist governments cannot be prevented from colluding to act even in mild crises. Governments with heterogeneous preferences can be induced to act in accordance with citizen preferences but only if they are jointly vulnerable electorally. It is only nationalist governments that can be relied upon to do what the citizens want them to despite the electoral vulnerability and the possibility for collusive agreements.
When the CPE does not exist, any equilibrium must involve some type of policy failure: either a false positive (governments intervene when they are not supposed to), or a false negative (governments do not intervene even when they are supposed to).
In the context of the EU, the “democratic deficit” is often alleged to arise from the Union being a “distant technocratic superstate run by powerful officials who collude with national governments to circumvent national political processes,” presumably with the end result being policies that the citizens do not want.Footnote 12 From this perspective, the most interesting false-positive failure is the one where the two governments agree to act in a mild crisis and share the policy burden. As the following proposition shows, one does not need the EU “superstate” to explain such outcomes: electorally minded national governments are perfectly capable of going against their citizens’ will without any further institutional obfuscation. The central result here is that electoral incentives could drive even nationalist governments to such hyperactive engagement but that the more electorally vulnerable the incumbent gets, the smaller the chances of such policy failure are.
Proposition 2
The following assessments constitute a false-positive burden-sharing equilibrium only if
$s \ge \bar s = \max (e_1, e_2 )$
:
-
• Each government acts regardless of the nature of the crisis.
-
• The citizens in each country reelect the incumbent when they observe multilateral action. When they observe any other outcome, they infer that the crisis is serious, reelect any government that acts, and replace any government that does not.
This equilibrium is collusion-proof and intuitive.
When citizens are quite certain that the crisis is serious, they are going to reward action and punish inaction even if they are still unsure about the precise nature of the crisis. Internationalist governments obviously benefit from this because they get to have their cake (they act) and eat it too (they get reelected) even though they are, in fact, acting against the wishes of the citizens when the crisis is mild. The electoral threat forces even nationalist governments to fall in line and participate when neither they nor, ironically, their citizens actually want to.
Citizens are, of course, quite aware that they might be precipitating the very behavior they are trying to prevent and they are willing to do so only if they believe that the probability of such a mistake is low. This is why a necessary condition for this equilibrium is for them to think that it is very likely that the crisis is serious and requires action (s is high enough). With such a belief they are willing to reelect their government even though there is a chance that it has acted contrary to their wishes. When the incumbent is more vulnerable electorally, their tolerance for such a mistake becomes lower (because the replacement they can elect is more attractive), which pushes the required initial beliefs further up.
False-positive failures are not restricted to burden-sharing arrangements. As the somewhat tedious analysis in Appendix A shows, when at least one of the governments is internationalist, equilibria with distributional conflict exist in which an internationalist government ends up paying the entire cost on its own (the burden-shifting equilibrium in Proposition B), or is at least forced to assume that burden disproportionately often (the limited burden-sharing equilibrium in Proposition C). Aside from showing that nationalist governments cannot be induced to carry more than their share, these cases do not add much of substantive significance to our analysis although the extreme burden-shifting scenario could be useful in understanding Slovakia's behavior (Appendix C).
It is the false-negative failure, however, that is of special relevance to the puzzle we set out to resolve, which is why we focus on it. We now investigate the possibility that governments do too little; namely, that they fail to act not only when the crisis is mild—as their citizens want them to—but also when the crisis is serious. This is a particularly egregious type of policy failure because it saddles the citizens with a deepening crisis that they will eventually have to pay to resolve. The central result is that electoral concerns could keep even internationalist governments from acting when the crisis is serious but the more vulnerable the incumbent, the less likely such policy failure becomes.
Proposition 3
The following assessments constitute a false-negative equilibrium only if
$s \le \underline s = \min (1 - e_1, 1 - e_2 )$
and
$w_i \le \bar w_i = \left[ {1 + t_i (1 - \alpha _i )C} \right]/\theta _i $
:
-
• No government acts regardless of the nature of the crisis.
-
• The citizens in each country reelect the incumbent when they observe inaction. When they observe any other outcome, they infer that the crisis is mild, reelect any government that does not act, and replace any government that does.
The equilibrium is collusion-proof but it is intuitive only for internationalist dyads.
${\rm \squf} $
This result should be jarring for it states that while internationalist dyads can experience this type of policy failure, dyads where at least one of the governments is nationalist cannot. To put it differently, it is only when both governments are internationalist—and thus very interested in acting regardless of the nature of the crisis—that a serious crisis might remain unattended with both governments remaining passive for electoral reasons. Ironically, this sort of massive policy failure that will saddle the hapless voters with the costs of a rescue from a wider and deeper crisis cannot occur when at least one of the governments is nationalist.
How do we explain this puzzling behavior? The answer lies in the underlying incentives of internationalist and nationalist governments. As long as it is rewarded for inaction, a nationalist government does not have an incentive to act when the crisis is mild even if doing so would also result in reelection. When voters observe such a government acting unexpectedly, they can safely infer that the crisis is serious, in which case they can also reelect it for doing the right thing, which, in turn, rationalizes its unexpected deviation. Unlike the nationalist government, an internationalist government cannot credibly signal that the crisis is serious in this way. If it expects to be rewarded for deviating, it will have an incentive to do so even if the crisis is mild, which means that when voters observe such a government acting unexpectedly, they cannot safely infer that the crisis is serious so they will not reelect the government. This, in turn, prevents the internationalist government from acting even in a serious crisis. In other words, since the internationalist government cannot credibly signal what it knows, the citizens cannot be induced to remove the electoral threat that is preventing the government from acting. Internationalist governments are prisoners of voter expectations: because they are known to want to do too much, they are condemned to do too little.
It is worth asking why this equilibrium is not susceptible to internationalist governments colluding to act even when they know that the crisis is serious. It is not really the threat to punish them both if they engage in multilateral action that prevents collusion. It is the lack of incentives to abide by the collusive agreement that is destroying its viability. In this equilibrium voters always reward the inaction of their own government regardless of what the other government does. This means that if governments agree to act in a serious crisis, each of them can do better by breaking their promise and doing nothing: whoever does this will both get reelected and saddle its erstwhile co-conspirator with the full cost of the action. The collusive agreement cannot be sustained, and internationalist governments end up doing nothing.
We now show how the model can rationalize Merkel's dithering strategy and explain both its sudden collapse and the electoral disaster that followed.
The German Politics of the Greek Bailout
The problems with Greece began in earnest shortly after the snap elections that brought a new socialist government to power in 2009. Prime Minister George Papandreou revealed that the previous governments had seriously mismanaged the economy and saddled the country with a crushing debt of 129.7 percent of GDP and a massive deficit of 12.7 percent of GDP. The debt was more than twice the size Eurozone members were allowed to incur, and the budget deficit was more than four times the agreed limits. The markets reacted immediately. Rating agencies began downgrading the Greek debt, and by the early spring of 2010, the government was effectively shut out of the international financial markets. Rumors about a potential agreement on a bailout for Greece spread through the Eurozone despite the clear “no bailout clause” in article 125 of the EU treaties. Any impetus for a concerted international action, however, foundered on Germany's stiff, if unexpected, opposition.
How are we to understand the German government's behavior? Scholars, politicians, and the media have advanced three explanations for Merkel's dithering. The first was a policy blunder: Merkel had made a huge mistake in believing that the crisis would not affect the Eurozone, and by the time markets proved her wrong, the crisis had nearly gotten out of hand. The second, argued by the chancellor herself (albeit only in retrospect), was that the delay was a strategy designed to coerce other governments to implement the right policies. These explanations are not consistent with the evidence during the critical months of March and April.
A third explanation turns on electoral motivations: Merkel tried to postpone what she knew would be a highly unpopular, but necessary, decision until after the elections in the country's most populous state, Nordrhein-Westfalen (NRW), on 9 May. These elections were critical to Merkel's governing coalition because a defeat for the Christian Democratic Union (CDU) in NRW would lead to a loss of control in the Bundesrat. This would jeopardize her government's plans for a radical overhaul of the tax and health systems, and an extension of the nuclear power program. These plans were opposed by the Social Democratic Party (SPD). Public opinion polls in NRW indicated a close race between the SPD and the CDU, and opinion poll experts predicted that the bailout debate could have a strong impact on voters.Footnote 13 These elections were so important that some analysts argued that all federal politics had come to a standstill because decisions had been either made or postponed because of them. Not only that, but NRW was “historically speaking, a seismograph for national politics.”Footnote 14
There was no shortage of speculation about an electoral motivation behind Merkel's delay, both in Germany and abroad.Footnote 15 The opposition was especially vocal in its allegations that a bailout was a foregone conclusion.Footnote 16 But it is one thing to assert that a political leader postponed the implementation of an unpopular decision until after an election, and it is quite another to explain why this strategy should work. How could citizens not see through such a transparent ploy? If a bailout was inevitable, putting it off would be, in the prescient words of the EU Green Party Leader Cohn-Bendit, “incomprehensible and politically very stupid.”Footnote 17 Our model can help explain why Merkel's electoral strategy made sense.
Equilibrium Selection
The first step in applying the model is to select among its several equilibria on the basis of the parameters necessary for their existence. From the vantage point of the German government, the situation between 11 January (when Eurostat officially questioned the Greek debt and deficit figures) and 27 April (when S&P downgraded Greek and Portuguese bonds) is consistent with parameter values that map onto the false-negative equilibrium. Recall that this equilibrium requires (1) an internationalist dyad, (2) citizens believing that the crisis does not require a bailout, and (3) costs of a serious crisis not being excessive.
First, given the express concerns of the other important Eurozone members and their ready willingness to participate in a common bailout early on, we can regard them as internationalist. Both the CDU and Angela Merkel were also regarded as internationalist. In fact, in party manifestos and expert evaluations, German governments tend to come out as more internationalist than other EU governments in general.Footnote 18 Merkel in particular had earned the nickname “Mrs. Europe” for her exceptional handling of the previously gridlocked negotiations for the 2007–13 financial framework.
Second, German voters did not believe that the Greek crisis was serious enough to affect their own well-being, and were consequently opposed to a bailout. Most of them believed that bailing out the Greeks was both unfair and unnecessary. While their chancellor was telling them that Greece would solve its own problems, the media was regaling them with stories of astounding Greek government largesse and endemic corruption.Footnote 19 The examples of this are too numerous to cite, but one egregious example provides a useful encapsulation of the issues and a glimpse at the tenor. On the day of Papandreou's 5 March visit to Berlin, Bild published an inflammatory “Dear prime minister” open letter full of assorted accusations:
If you're reading this, you've entered a country different from yours. You're in Germany. Here, people work until they are 67 and there is no 14th-month salary for civil servants. Here, nobody needs to pay a €1,000 bribe to get a hospital bed in time. Our petrol stations have cash registers, taxi drivers give receipts and farmers don't swindle EU subsidies with millions of non-existent olive trees. Germany also has high debts but we can settle them. That's because we get up early and work all day. We want to be friends with the Greeks. That's why since joining the euro, Germany has given your country €50 billion.Footnote 20
Given these sentiments, most Germans instinctively approved of the schwäbische Hausfrau strategy that Merkel had debuted in 2008 when she warned that doling out credit to rescue the American finance sector would exacerbate the meltdown caused by the bursting of the real estate bubble. For wide swaths of the population, fear of inflation and aversion to debt had become part of a culture that emphasized frugality and solvency.Footnote 21 Many Germans believed that a bailout would endanger the stability of the Euro rather than support it.
It should, therefore, come as no surprise that the Germans were dead set against a bailout, in part because of austerity measures that had been necessary to meet fiscal consolidation targets in Germany.Footnote 22 Polls consistently showed that only 20 to 25 percent supported helping Greece, and Germans tended to be distant outliers compared to other Europeans on the causes and consequences of the crisis.Footnote 23 In March, an IFOP survey reported that 78 percent of Germans believed that the Greek government was responsible for the crisis rather that it being part of a global crisis or a result of financial speculation. The average of those who shared that sentiment among those surveyed in Spain, France, Italy, and the UK was only 54 percent. The majority of Germans also did not think that the crisis was significant either personally or to those around them: 55 percent compared to an average of 36 percent among the other Europeans. Germans were also far more confident that their country could not suffer the same fate as Greece: 66 percent compared to an average of only 41 percent for the others. Since they blamed the Greeks for the crisis and did not believe it would affect them, 76 percent did not want to help Greece. Majorities in Italy (67%), Spain (55%), and France (53%) thought that their governments should help Greece in the interests of European solidarity. The only citizens the Germans resembled in their hawkishness on the bailout were the British (78% opposed), but the UK was not a member of the Eurozone.Footnote 24 In fact, about a third of the Germans would rather see Greece expelled from the Eurozone than pay to bail out its government, and in this sentiment they again exceeded everyone else.Footnote 25
Third, the costs of continuing a serious crisis were not seen as excessive by political elites and publics alike. By March, the other Eurozone members and the IMF had reached a consensus that the crisis was serious, but in their initial bailout agreement from 11 April they estimated that only about EUR45 billion in loans would be sufficient to rescue Greece. The EUR15 billion IMF share was comparable to its loans to Brazil 1999 and Mexico in 1994, and the overall package was akin to the bailout for Argentina in 2001. In other words, while the crisis was clearly serious from a Eurozone perspective, it was perceived as manageable. The economic costs were also not expected to be grievous—the Greeks did not even request the activation of the emergency loans under this agreement until 23 April, and the credit ratings on government bonds in Greece itself but also in Portugal, Ireland, Italy, and Spain (the PIIGS countries where the crisis was most likely to spill into) remained at investment-grade levels until 27 or 28 April.
The Schwäbische Hausfrau Policy
In line with the equilibrium logic, Merkel adopted a laissez-faire policy. This position was not difficult to sustain in the early months while the crisis seemed localized and within Athens’ ability to stem. Members of the government, the coalition parties, and leading newspapers all insisted that Greece should cope alone.Footnote 26 The EU Council meeting on 11 February limited itself to assurances of political support for Greek reforms while emphasizing the need to abide by the rules. Even when the situation in Greece took a turn for the worse amid nationwide protests against the austerity program of 5 March, the Eurogroup refused to commit to any financial help and instead pressed for further austerity measures.
In exasperation, Papandreou warned that Greece might have no choice but to turn to the IMF for help if the Eurogroup did not put together a rescue package at the EU summit scheduled for 25 March. His particular concern was that the waffling EU response had fanned the flames of speculation, causing Greek bond yields to top 6 percent. At such an exorbitant rate, Athens had no hope of financing itself via the markets out of the crisis. The only way to stop the betting against Greek debt was through a firm commitment to a bailout by the Eurogroup or, failing that, assurances of loans from the IMF.Footnote 27 Everyone—markets, Eurogroup finance ministers, the head of the OECD, and the president of the European Commission—agreed with him. Everyone, that is, except the Germans.Footnote 28
Reflecting both the moral hazard perspective and the widespread popular opposition to a bailout, Merkel told the Bundestag on 17 March that rushing aid to Greece in “a quick act of solidarity” was wrong, and that a fundamental solution has to be devised; a solution that would allow for the expulsion from the eurozone of countries that persistently break its financial rules.Footnote 29 When the inevitable hue and cry arose over breeching the expulsion taboo, Merkel reminded everyone that Greece had yet to ask for financial aid, insisted that she did not believe the country was facing imminent insolvency, and flatly stated that any discussion of a bailout was off the table for the upcoming EU summit.Footnote 30
Consequently, the statement released at the 25 March summit harped, much like its 11 February predecessor, on the need to follow the rules, but went further by promising “a package involving substantial IMF financing and a majority of European financing.”Footnote 31 This seemed to have committed the Eurogroup to a bailout and satisfied Germany's demand to get the IMF involved. On the other hand, the statement also insisted that since Greece had not requested any financial help, the rescue mechanism was not being activated. It also emphasized that the loans would be at nonconcessionary rates, that they would be provided only as an absolutely last resort, and that their provision would require the unanimous consent of the euro area members after assessments by the Commission and the Central Bank.Footnote 32 The official statement from the Chancellor's Office, however, chose to emphasize just how hedged that promise was. In only thirteen sentences of text, it managed to say that the package was a “last resort,” “very last resort,” and “absolutely last resort.” After professing a commitment to the common currency, it clarified that any disbursements would involve “strict criteria” and had to be “authorized unanimously,” and that the loans would be priced “in line with the de facto risks.”Footnote 33
Merkel's tough talk on Greece brought her political gains domestically. Figure 1 shows that initial rumors of a bailout at the end of 2009 led to declining support for Merkel. However, after her staunch opposition to the Greek bailout, support increased and stabilized in March and April of 2010. We can see the same, even slightly stronger, pattern in support for the CDU. This period also saw a stabilization in the share of voters who believed that the CDU government had competently handled the economy.
Figure 1: Public opinion in Germany during the Eurozone crisis.
Fiddling While Rome Burns
The financial support mechanism that the Eurozone heads of state had committed to on 25 March became fully operational on 11 April when the finance ministers provided the details along with the requirement that Greece implement further austerity measures and report frequently on the status of their implementation. Athens immediately began negotiations on the extent and severity of these additional measures. On 22 April, Eurostat revised Greece's estimated deficit to 13.6 percent of GDP (up from 12.7 percent).Footnote 34 This caused Moody's to cut Greek bond rating to A3, citing “significant risk” and warning that the rating would slide further “unless the government's actions can restore confidence in the markets and counteract the prevailing headwinds of high interest rates and low growth.”Footnote 35 The ten-year bond yield surged to an astonishing 8.8 percent, and the spread from Germany's bond widened by 5.75 percent. Schäuble still clung to the established narrative, claiming in an interview in Deutschlandfunk the very same day that the Greeks would not ask for help for weeks, perhaps until mid May.Footnote 36 The Greek government formally requested financial assistance under the new mechanism on the following day.
By the end of April, the economic and financial situation in Greece had worsened so much that experts no longer thought that the bailout package—even if were to come—would suffice to stem the crisis. Greece's debt had reached almost EUR300 billion, and after the 4 percent interest rate hike, its borrowing costs were 67 percent higher than they had been in February 2010.Footnote 37 It was unlikely that Greece would be able to service the EUR8.2 billion that were about to mature on 19 May at such prohibitive rates.Footnote 38 With the country headed toward almost certain default and financial markets in turmoil, experts predicted that a restructuring of Greek sovereign debt was unavoidable although Schäuble denied it.Footnote 39 The yields on two-year Greek government bonds had increased to over 13 percent: it was now safer to lend money to Iraq or Venezuela than to Greece.Footnote 40
On the Road to Damascus
In this heated atmosphere, Standard and Poor's (S&P) 27 April downgrade of Greek government debt to junk (BB+ for long-term and B for short-term bonds) and Portugal's to low investment grade (A−, closing on the territory previously occupied by the Greek bonds) unleashed a veritable panic. Because the downgrade was accompanied by a warning that the agency expected investors to lose between 50 and 75 percent if Greece defaulted, the fallout was immediate and severe.Footnote 41 European stock markets plummeted as investors voiced fears over the crisis and the risk of contagion.Footnote 42 On 28 April, S&P downgraded the Spanish long-term debt to AA, and an Italian bond issue failed to garner expected support. The borrowing costs for Ireland, Italy, and Portugal climbed as experts became increasingly convinced that a Greek default would unleash a series of defaults in the other PIIGS countries.Footnote 43 The crisis threatened to engulf the entire Eurozone, not just its weakest members. Sales of the euro accelerated, leading the common currency to plunge to its lowest value against the dollar in over a year and, since the yuan was tracking the dollar, against the Chinese currency as well.Footnote 44
The heads of the IMF and the ECB turned the screws on Germany to act, emphasizing the “absolute necessity to decide very rapidly” and “to act swiftly and strongly.”Footnote 45 Astonishingly, even now Merkel insisted that Greece had to implement an “ambitious” austerity program, and while she believed that the negotiations had to be “accelerated,” it was only after they had concluded that Germany would “make its decisions” on whether to grant aid.Footnote 46 A source close to the EU Spanish presidency indicated that the summit to discuss aid would be held on 10 May, a day after the NRW vote.Footnote 47 The peculiar scheduling would not be surprising to anyone who was aware that the latest polls found 57 percent of Germans adamantly opposed to a bailout (and only 33 percent in favor).Footnote 48
With the ship rapidly sinking, however, political action was unavoidable. On 3 May, the German government introduced the “Act on Financial Stability within the Monetary Union” that would clear the way for Germany's contribution to the bailout. The act passed on 7 May after heated debate, and became effective on the following day. On 9 May, the EU finance ministers assembled for an emergency meeting approved the rescue package totaling EUR500 billion, of which Germany's guarantees were EUR123 billion (with a possible additional EUR24.6 billion). The IMF also approved Greece's request for a Stand-by Arrangement of EUR30 billion, with an immediate release of its first tranche of EUR5.5 billion to refinance the Greek bonds maturing in ten days.
The Equilibrium Cost Condition Violated
The false-negative equilibrium can rationalize Merkel's opposition to a bailout despite her knowledge that the crisis was serious. However, the strategy required her to delay all the way until after the elections and she did not. The German bailout agreement passed in the Bundestag two days before the elections in NRW. These elections were an unmitigated disaster for the CDU, which lost by 10.2 percent relative to its 2005 performance, making this its worst electoral defeat in NRW ever. The government was replaced with a coalition of SPD and Greens, and Merkel lost the majority in the Bundesrat.Footnote 49 On 10 May, Merkel announced that the long-promised tax cuts were off the table for at least two years, and on the following day the German cabinet approved EUR123 billion for the rescue fund. The media erupted with outrage. Bild screamed, “Yet again, we are the idiots of Europe” for paying so much for “bankrupt neighbors” without money for tax cuts at home.Footnote 50 The political ramifications of the NRW loss were not merely temporary setbacks; they proved as costly and persistent as the gloomy forecasts had predicted. As Figure 1 illustrates, public support for Merkel fell by more than 18 percent to an all-time low, and support for the CDU fell to a low of 31 percent. Support for Merkel would not recover to the (uncharacteristically low) levels of the immediate pre-crisis months for two years, and support for the CDU would take even longer.
But if Merkel's dilatory tactics were motivated by domestic political considerations, why did she reverse course when she did, and why did she fail to persuade voters that this had been the right decision?
To understand the abrupt volte-face of 2 May, we need to recall that one of the necessary conditions for this equilibrium is that the expected costs of a serious crisis that is allowed to deepen are not excessive. When this condition is not met, then the government will have an incentive to deviate in a serious crisis and agree to a bailout even if doing so would cost it the elections. The unexpected downgrades on 27 and 28 April, with their devastating implications for the Eurozone, were catalytic. They convinced Merkel not only that the costs of the crisis would be significantly worse than expected but that the situation was deteriorating much more rapidly than she had anticipated. This made further delay tantamount to permitting the Eurozone to go to ruin. Merkel's original dilatory strategy was thus no longer optimal.
It is crucial to realize that Merkel's tactic was predicated on there being no drastic changes in Greece's position. Had the downgrade been anticipated, it would have been incorporated into the expectations, and the false-negative equilibrium would have been unsustainable, implying no delay for the bailout. The S&P actions, however, caught everyone by surprise. The IMF chief went so far as to say that the rating agencies should not be “believed too much.”Footnote 51 Merkel promised to “press for the creation of a ratings agency in Europe so that European financial markets become more stable and reactive.”Footnote 52 As this official annoyance at S&P's actions shows, policymakers had not anticipated the downgrade.Footnote 53
The unexpected downgrade put the chancellor in a quandary. She had spent the last few months telling the Germans that the Greek crisis was not their problem, that the Greeks had to get their act together, and that German taxpayers would not be held liable for the excesses of the Greek government. By all accounts, she had succeeded marvelously. The problem Merkel now confronted was that while she was convinced that the crisis was serious for Germany, the voters clung to their original beliefs.Footnote 54 With these beliefs, they would treat a bailout as a deviation and punish it accordingly. The only way to avoid this would be to persuade them to revise their beliefs. Given the parameter configuration (all else equal except much higher costs, w i ), if voters were to believe that the crisis is serious with a higher probability, s, the equilibrium would be the burden-sharing one in which the governments act and get reelected.
The Unsuccessful Attempt to Coordinate on a New Equilibrium
With everything that was at stake domestically, Merkel tried very hard to persuade German voters that the bailout was crucial for the German economy. The German government switched to damage-control mode almost immediately after the second downgrade. Schäuble now insisted that loans for Greece were good for Germany.Footnote 55 Merkel doubled down and said
It is about nothing more and nothing less than the future of Europe, and therefore the future of Germany in Europe … A good European is not necessarily the one who helps quickly. A good European is the one who respects the European treaties and the relevant national law, and helps accordingly to ensure the stability of the Eurozone.Footnote 56
She went on a veritable media blitz with news conferences and interviews on the day the Eurozone members approved the bailout package.Footnote 57 She made fifteen personal appearances in NRW alone and spent the week before the election giving numerous interviews on television.Footnote 58
The voters were not buying it. Since June 2009, the fraction of Germans who thought that the current economic situation was good or very good had been steadily increasing. The same trend obtained for the expectations about the future.Footnote 59 Compared to January 2010, when 64 percent of Germans thought that the worst of the crisis was still to come, by May only 56 percent thought so.Footnote 60 In mid April, 78 percent of Germans believed that their own economic situation would either not be affected by the crisis or improve over the next few years; 59 percent believed that unemployment would either remain stable or decline; and 71 percent believed that the economy would either remain as is or improve.Footnote 61 Even after the bailout 56 percent of Germans continued to believe that aid to Greece was wrong (only 39 percent were in favor), and that despite 67 percent thinking that the euro would destabilize over the next year.Footnote 62
Recognizing the inherent weakness of the chancellor's new position, the opposition now pounced on it, making it the most important topic in the electoral campaign in NRW.Footnote 63 As Klaus-Peter Schöppner, head of the polling institute Emnid, said, “The issue has electrified people as seldom before and is going to play a determining role” in the election.Footnote 64 The last poll published by Bild on the eve of the elections showed that 20 percent of NRW voters said that the bailout would affect their decision.Footnote 65
At the end of the day, German voters had no reason—in evidence or logic—to believe Merkel's sudden conversion. And so they did not, treating the bailout as a deviation in the false-negative equilibrium that required a punishment at the polls.
Two Alternative Explanations
We have now made the case that Merkel's decisions during the Greek debt crisis were motivated by electoral considerations. There are, however, two common alternative explanations of her behavior that we would like to address.
A Policy Blunder?
One possible explanation interprets the delay as a failure of German politicians to see past the cultural and ideological commitment to austerity, and a failure to understand how financial markets could spread the Greek malady to other vulnerable members of the Eurozone. Whereas the cultural affinity to austerity policies and the popular fear of inflation certainly did not make it easier for the German government to commit to a bailout, there are two problems with this explanation.
First, it requires one to maintain that Merkel had been singularly deluded when other governments, the EU Commission, and the IMF were all in agreement that the Greeks needed a bailout. European leaders urged Merkel not to delay the bailout to Greece, but to act in solidarity with other members of the Eurozone. It is difficult to see how Merkel and her ministers could have been so out of touch with market reality, especially in late April when they still maintained that Germany could refuse to aid Greece.
Moreover, if the German government did not care about the Greeks, it presumably did care about the investments of German banks whose exposure to Greece in the first quarter of 2010 was, at USD44.2 billion (24 percent of the total exposure of European banks), second only to France's USD71.1 billion.Footnote 66 As Alessandro Leipold, former acting director of the IMF European department, noted, there were “intrinsically strong German interests” at stake.Footnote 67 There is no doubt that the German government was aware of these highly risky entanglements.Footnote 68 It is very implausible that it would not have acted upon this knowledge to prevent an almost certain spillover of the crisis to Germany just because of its cultural commitment to austerity, especially since this would have almost inevitably created the inflationary pressures that the government was determined to prevent.
Second, and crucially, the explanation cannot account for the clobbering the voters in NRW delivered to Merkel's party. Suppose that the chancellor had been just as convinced as the voters of the wisdom of the schwäbische Hausfrau strategy until the end of April but then underwent a rapid conversion. Why would the voters not follow her in that? After all, she had been the most hawkish Eurozone leader on Greece, and if she had suddenly come to the realization that a bailout was necessary to save the euro, the voters should have believed her. Only Nixon could go to China, and only Merkel could go to Greece. But the voters did not believe her.
One might be tempted to argue that the German voters punished the CDU because Merkel was inconsistent—first opposing the bailout, but then flip-flopping—or because her Machiavellian tactics had worsened the crisis, saddling Germany with six times the costs. But this was not how the Germans voters interpreted it. They remained unconvinced about the seriousness of the crisis. Polls in late April and early May showed that the majority of Germans opposed the bailout because they believed it was wrong to aid Greece. Surveys also revealed that they did not consider the crisis a top priority for Germany, and did not expect it to affect them adversely personally. These data point to a failure to carry the voters on the new policy, not to a punishment for not dealing with a serious crisis promptly.
A War of Attrition?
A second possible explanation centers on a potential distributional conflict as the source for the delay. Accordingly, Merkel was holding out for better terms, both from the Greek government and from the fellow Eurozone members. The Greeks had to commit to even more drastic austerity measures, while the Eurozone members had to agree to terms that would not prejudice the credibility of the threat to let future spendthrifts sort out their own problems. The peculiar insistence on IMF participation—long opposed by other Eurozone members—must be seen in that light because that organization had a lot of experience of imposing unpopular reforms on recipient countries.
The war-of-attrition logic can explain the initial phase of negotiations. As the crisis worsened in early 2010, Greece was increasingly willing to accept tougher austerity measures as demanded by the German government. The problem with this explanation is that Germany had already achieved all of its stated goals in principle with the 25 March agreement, and in practice with the 11 April decision to make the rescue mechanism fully operational.Footnote 69 The IMF had been involved since the March agreement, and Schäuble himself had indicated in an interview that the outline of the austerity program had been decided in mid March.Footnote 70 Moreover, Merkel had already dropped the insistence on market rates for the loans in the 11 April agreement. Finally, the “shallow text” of the 11 February agreement, with its emphasis on the defense of the monetary union's stability, was a “crucial strategic coup for Merkel” because it allowed her to deflect a potential bailout challenge by the Constitutional Court.Footnote 71
While it is true that the Greek government announced a third wave of cuts in conjunction with the 2 May deal, one cannot argue that Merkel had delayed obtaining its formal commitment. The chancellor herself claimed to have done so because without Athens announcing new austerity measures, giving aid “would have had the opposite effect” to calming markets.Footnote 72 The irony of this statement in light of the reason the markets had gone berserk cannot be overstated.
In fact, it was because of this that the press and the opposition had speculated that Merkel's tough line had been a domestic kabuki theater at least since March.Footnote 73 That is also why Steinmeier accused Merkel of playing a double game between Brussels and Berlin, “Madame No—that was a huge hoax.”Footnote 74 He also dismissed the notion that the delay had been a part of some coherent plan to create a better policy. As he told Merkel, “You drifted around like a windsock. Then in retrospect you call that your strategy. Your double game has cost us an enormous amount of trust and respect in Europe.”Footnote 75 The allegation of ex post rationalization is also supported by the fact that the German government only belatedly (after 28 April) started to insist on the importance of the crisis for Germany itself.
This explanation also has a flaw in the logic of the strategy itself given that it was being played in an electoral shadow. Merkel could have denied that the Greek crisis posed a problem for Germany in an attempt to signal that her government had little incentive to act unless all its stringent conditions had been met. This might have increased the credibility of the threat, but since she had done so publicly, it also signaled to the German electorate that a bailout was unnecessary. Judging from the opinion surveys and the prevalent opinion in the press, the voters seem to have believed her. But if Merkel knew a bailout was coming and was merely stalling for terms, this would have been a silly thing to do because the bailout would certainly upset the voters. A more profitable strategy would have been to indicate that a bailout was necessary and outline the conditions Athens had to satisfy to obtain it. Of course, Merkel later claimed that this had been precisely what she had done, except that somehow nobody had understood her that way: not the IMF, not her fellow European heads of state, not the domestic opposition, not the press, and not the voters.
Conclusion
We wondered why cooperation on the first Eurozone bailout was so difficult to achieve despite strong pressures to do so. In particular, we were puzzled by the fact that the most important holdout was the country that stood to lose most if the crisis got out of hand. Our answer is that intergovernmental actions take place in front of domestic audiences who might use the actions they can observe to form opinions about their appropriateness and condition their electoral choices on the inferences they make. Strong domestic opinions can lead to suboptimal foreign policies although the threat of electoral sanction and the promise of electoral reward can also induce governments to cooperate only when their citizens want them to.
Our model has broader implications for international relations theory. Consider the interaction between governments and their citizens. By bringing in the electoral motivation, we immediately raise the familiar principal-agent problem of how citizens can get governments with divergent preferences to behave.Footnote 76 The context we study, however, is novel because we incorporate a crucial feature of international politics: the presence of other governments who are agents of different principals but whose actions are observable and therefore potentially informative as well.
Although this setting will be familiar to anyone who studies two-level games, our emphasis on signaling (as opposed to distributive conflict) is new. Consider the difference between a mixed and an internationalist dyad. Whereas the presence of the nationalist government ensures that signaling will be credible and as a result international action will not occur in a mild crisis, a combination of internationalist preferences ensures that signaling will be unreliable: these governments would collude and as a result the CPE cannot be sustained (Propositon 1). The “democratic deficit” can occur because the lack of transparency in international negotiations provides governments with opportunities to collude in pursuit of their preferences to the disadvantage of the voters. In this context, international cooperation can become domestically abusive.
The electoral control mechanism can be further weakened by the beliefs of the citizens themselves: whenever they hold strong priors about the desirability of some particular foreign policy, governments might not be able to signal the need for a different policy even when this need might be real and the citizens would want to know it. As Propositions 2 and 3 show, such circumstances can produce various international cooperative behaviors that fail the domestic normative test. Government efforts to influence citizen beliefs can become self-fulfilling prophecies and, as the German example demonstrates, turn into serious obstacles to implementing policies the voters would actually prefer.
The model also yields a different take on how the diversity of preferences among members of an international organization affects the prospects for cooperation. It is commonly accepted in the literature that the more heterogenous the membership, the “shallower” the cooperation.Footnote 77 When scholars argue to the contrary, they point to heterogeneity increasing the opportunities for issue linkages and coalition formation.Footnote 78 In contrast, we find that heterogeneity can have a positive effect on the prospects for international cooperation because the presence of diverse governments can enable credible information transmission to the voters. In an environment plagued by informational asymmetries credible signaling by governments can be crucial in securing their cooperation on international issues by helping them avoid adverse domestic reactions to such behavior. When it comes to working agreements, the breadth and the depth might be mutually reinforcing.
It is uncontroversial that domestic politics matter for foreign policy. Far more important is how they matter, and here there has been a distinct tendency to use domestic-political arguments to explain why states choose foreign policies that are suboptimal from some normative perspective. Our approach explicitly rejects the notion of a normative standard that is defined without reference to citizens’ preferences. “International cooperation” must be understood not in terms of whether governments abide by their agreements or agree to bear costs when benefits diffuse to other governments, but also in relation to the domestic preferences these governments are supposed to represent.
This normative perspective allows us to go beyond treating domestic politics as a foil for foreign policy or a last-resort explanation of some shortcoming it is supposed to have, and explore how foreign policy actions—cooperative or not—can inform citizens and perhaps enable them to implement electoral strategies that provide incentives to their government to choose policies in line with their preferences. In this light, our finding that the CPE can be supported in mixed or internationalist dyads only when the governments are jointly vulnerable electorally is illustrative as an instance of international cooperation that produces outcomes to the citizens’ liking that would not be achievable by governments that are unconstrained domestically. “Domestic politics” need not be dirty words when it comes to foreign policy.
Supplementary Material
Supplementary material for this article is available at <https://doi.org/10.1017/S0020818317000406>.
