The Political Economy of the Greek Debt Crisis: A Tale of Two Bailouts
Silvia Ardagna and Francesco Caselli First draft: February 2012; Final version: January 2014

Abstract We review the events that led to the May 2010 and July 2011 bailout agreements. We interpret the bailouts as outcomes of political-economy equilibria. We argue that these equilibria were likely not on the Pareto frontier, and sketch political-economy arguments for why collective policy making in the Euro area may lead to suboptimal outcomes.

Most modern sovereign debt crises have been managed in Washington, DC, through the combined e¤orts of the International Monetary Fund (IMF) and the US government. A distinctive feature of the crisis that has engulfed European sovereign-debt markets since the fall of 2009 has been that the IMF has played only a supporting (albeit important) role, while the management of the crisis has been driven by European institutions: the council of …nance ministers (ECOFIN), the European Council (EC, made up by all the heads of government of the European Union) and the European Central Bank (ECB). To the extent that the IMF is largely a technocratic institution (though of course not entirely immune from political in‡ uence) while ECOFIN and the EC are made up of politicians, one may expect the management of the crisis by the EC to be more a¤ected by electoral concerns. Furthermore, since there are 27 members to the EC, representing countries with potentially di¤erent interests, one may expect that bargaining and compromise will play a greater role than in cases where the two players are simply the
Ardagna: Goldman Sachs; Peterborough Court 133 Fleet Street, London EC4A 2BB; silvia.ardagna@gs.com. Caselli: LSE, CFM, CEP, CEPR, and NBER; London School of Economics, Houghton Street, W2A 2AE, UK; f.caselli@lse.ac.uk. We bene…ted from insightful conversations with Wendy Carlin, Michael McMahon, Herakles Polemarchakis, Morten Ravn, Ronny Razin, and Silvana Tenreyro, and constructive comments from Steve Davis and four referees. Excellent research assistance from Michael Conroy, Kilian Huber, and Dana Kassem is also gratefully acknowledged.

1

IMF and the country whose debt is under pressure. Finally, the presence of an additional powerful player, the ECB, whose actions can greatly a¤ect the outcome of the crisis, may be expected to impact both the incentives and the constraints of the EC, and introduce further di¤erences in its policy response. In this paper, we revisit two salient passages of the crisis, keeping an eye out for indications that the considerations above have played a role in shaping the policy response. We do this with two modest goals in mind. First, we assess the extent to which some of the observed decisions can be rationalized as political-economy equilibrium outcomes of the complex game brie‡ sketched out above. Second, we y discuss the extent to which such political-economy equilibria appear to be e¢ cient and, if they are not, whether the particularities listed in the previous paragraph help us understand why the policy response has been inside the Pareto frontier. Because the ongoing crisis has engulfed many countries and the policy response has been complex and multidimensional, we have to limit our overview both in time and space. We have therefore decided to focus narrowly on policies towards Greece, where the crisis started, and to the period between September 2009 to July 2011. In practice, this takes us from the inception of the crisis, through the period leading up to the …rst Greek bailout (May 2010), and all the way to the second bailout at the end of our period
1 of analysis. Hence, the paper largely turns into “a tale of two bailouts.”

The paper is a …rst attempt to organize ideas about events that are exceptionally recent and “raw.” Our comments are therefore speculative at best. With that caveat, we reach the following tentative conclusions. Regarding the May 2010 bailout, we argue that this was a political-economy equilibrium in the sense that, for each of its signatories, it was individually rational to agree to it. In particular, the deal on o¤er was preferable to unilateral and disorderly Greek default, which would have been the outcome had any individual player refused to sign up. However, this does not necessarily mean that the agreement of May 2010 was the best possible deal available to European leaders, or that it was collectively rational. We speculate that a programme more generous towards Greece, particularly one that gave Greece more time before having to return to borrowing on private markets, might have made everyone better o¤. Concerning the July 2011 bailout, we reach similar conclusions, except that the Pareto inferiority of the deal seems even more apparent. In particular, in addition to unrealistic goals for de…cit and debt reduction and access to private lending, that agreement required that Greece seek a symbolic negotiated reduction in the value of its outstanding debt to private creditors (known as “private sector participation,”
1 For a much more encompassing view of the European sovereign debt crisis see, e.g., Lane (2012). For an analysis of policy towards Greece subsequent to the July 2011 bailout, see, e.g., Zettelmeyer, Trebesch, and Gulati (2013).

2

or PSI). We argue that this haircut component was not helpful in solving Greece’ problems, while at the s same time complicated the prospects of other peripheral Eurozone countries engulfed in contagion. It would thus have been Pareto superior to scrap it altogether. Having argued for the potential Pareto inferiority of both bailouts, we then use the three above-named features of domestic electoral concerns, bargaining among several parties, and presence of the ECB as an additional player to sketch arguments for why suboptimal equilibria can arise in the novel circumstances of the European debt crisis. We brie‡ consider the possibility that the main signatories had asymmetric y beliefs about the prospects of successful Greek stabilization under the terms of the bailouts. We also discuss at greater length the role of voters’beliefs and voter intransigence in some of the core countries. We …nd neither of these explanations entirely satisfactory. Hence, we propose two additional political frictions that appear to be particularly relevant in the context of EC decision-making. The …rst one centres on a communication friction between political leader and voters, which distorts the bargaining stance of the leader vis-a-vis its European counterparts. The second is a bargaining friction that is linked to the time-limited nature of bargaining sessions in EC summits. The paper is organized as follows. Section I looks at the period leading up to the …rst bailout, and assesses the …rst bailout as a political economy equilibrium. Section II does the same for the period up to and including the second bailout. Section III sketches arguments for why bailout agreements reached within the EC may be inside the Pareto frontier. Section IV summarizes and concludes. I. The First Bailout I.A The Road to the May 2010 Bailout The proximate trigger for the loss of market con…dence in Greece’ debt was the October 2009 ans nouncement, by a newly-elected government, that the overall budget de…cit was much larger than stated by the outgoing one. Instead of 6 to 8% of GDP, the de…cit was now deemed to be between 12 and 13%. With the debt/GDP ratio at 115%, and mediocre growth prospects, the announcement led markets to question the long-run solvency of Greece. Such concerns came to a head in December with the …rst of what will prove to be a long series of credit-rating downgrades. In a mechanism often seen during the crisis (and previously during the crisis in the …nancial sector in 2007-2009) market views on Greece changed astonishingly rapidly, in seemingly self-reinforcing fashion. By the spring Greece, which only six months before could borrow at rates essentially identical to those paid by Germany, was e¤ectively shut out of the …nancial market. The Greek government’ response to this predicament was twofold. On the domestic front, it announced s 3

and implemented a number of austerity measures, aiming to reduce the budget de…cit (see the Online Appendix for a partial list). These measures were aggressive, but still insu¢ cient to …ll the vast hole that had opened in the budget, and that markets had become unwilling to …ll. On the external front, therefore, the government begun exploring options for a bail out that would allow it to avoid default. The agreement that was struck by the European Council and the IMF in May 2010 amounted to a fairly straightforward “division of labor”where Greece committed to a severe austerity programme in exchange for a signi…cant amount of o¢ cial …nancial assistance.2 Speci…cally, Greece committed to bringing the de…cit down to 3% of GDP by 2014, with detailed quarterly targets, the compliance with which was to be monitored by o¢ cials from the IMF, the European Commission, and the ECB. All other Euro area states were to make bilateral loans to Greece, roughly in proportion to the size of their economies, for a total of approximately £ 80bn over three years. The IMF was to lend an additional £ 30bn over the same period. The role of the bailout funds was to …ll the funding gap left by the austerity programme, for the period deemed necessary before Greece could return to borrow on …nancial markets at acceptable terms. Table 1 Here Some key elements of the May 2010 plan are shown in Table 1. The plan called for Greece’ large s primary de…cit to disappear by 2012, and turn into a signi…cant surplus by the end of the programme. Overall de…cit reductions were signi…cantly front-loaded, presumably in order to establish credibility but also, of course, to reduce the needed external assistance. This plan was expected to lead the debt/GDP ratio to stabilize by 2013. Clearly, any such expectation was contingent on assumptions about the future path of GDP, which are also shown in the table. Greece was expected to go through a severe recession lasting until 2012, with sub-standard growth in that year as well. Perhaps more importantly, the adequacy of the bailout to …ll the funding gap depended not only on Greece’ meeting its de…cit targets, but also s on it recovering access to credit from private agents in a relatively short time. The last three columns of Table 1 show the assumed path of interest-rate spreads, as well as planned issuance of short- and medium/long-term debt. Although the latest tranche of aid was scheduled to be paid in April 2013, the program assumed that Greece would have been able to issue EUR 4bn in medium and long-term bonds in 2011, EUR 23.4bn in 2012, and EUR 34.9bn in 2013. The bailout was received with considerable scepticism by economic and …nancial commentators, as documented in our review of the …nancial press in the Online Appendix. Market reaction was also initially
2 The

agreement was reached by ECOFIN on May 2 and endorsed by the EC on May 7.

4

negative, with spreads on Greek bonds surging between May 2 and May 9, as also documented in the Online Appendix. These negative responses appear justi…ed by our own rough stab at assessing the feasibility of the programme. In the Appendix we compare the May 2010 stabilization programme to the largest episodes of …scal adjustment in the OECD since the 1980s. Brie‡ if one looks only at the percentagey, point reduction in the primary de…cit called for by the agreement, the e¤ort requested of Greece was not unprecedented. However, the macroeconomic backdrop was exceptionally adverse, with much lower predicted GDP growth and much worse initial cyclical position than in any major prior adjustment, and without any realistic prospects of the kind of signi…cant real depreciation that have arguably assisted in other cases.3 If the prospects for success of the May stabilization programme were so poor, why did European leaders agree to it? And were there possible alternatives that would have made all participants better o¤? The next two subsections tackle these questions in turn. I.B Understanding the Bailout In this subsection we take the provisions of the May agreement as given, and we ask whether it was individually rational for the main parties to agree to it – conditional on a plausible description of what refusing to sign up would entail. Before engaging in this analysis we need to brie‡ discuss the objective functions of the participants in y the agreement (we return to this issue in subsequent sections). As all the key players in the agreement were elected heads of state/government, we assume that objective functions are increasing in two arguments: the probability of reelection and a measure of social welfare.4 Hence, to discuss the costs and bene…ts from a deal for the key players we need to identify both the economic and the electoral implications of di¤erent choices. Individual rationality for the Greek leadership Since Greece was shut out of the …nancial markets, and running a de…cit in excess of 10% of GDP, it would certainly have defaulted on its debt had it refused the bailout on o¤er. The consequences of a possible default were deemed (correctly, in our view) to be catastrophic for Greece. It is highly unlikely that a default would have restored market access, meaning that Greece would still have had to incur a gigantic and instantaneous …scal adjustment of a magnitude at least equal to the primary de…cit –in the
Perotti (2011) on the importance of real exchange rate depreciation for the success of previous …scal adjustments. that social welfare here is intended very broadly, and, in particular, may not refer exclusively to the welfare of the citizens of one’ own country. For example, hypothetically there may be heads of government that feel such a very strong s European identity that they may be willing to trade-o¤ one’ own countrymen’ welfare for the greater good of Europe. s s
4 Note 3 See

5

order of 8% of GDP in mid-2010. Furthermore a Greek sovereign default would have been very costly for the Greek banking and …nancial sectors. In Q1 2010, Greek banks’ exposure to the general government amounted to 11% of their …nancial assets, and the equivalent …gure for insurance companies and pensions funds was 29%. Needless to say, the Greek government would have been incapable of recapitalizing these institutions in case of default. Hence, both because of the brutal …scal retrenchment that a default would have required, and because of the collapse of the banking and …nance sector, Greece was likely to experience a deep depression following a default.5 Furthermore such economic dislocation would likely have spelled political death for the prima minister (and possibly a long exclusion from power for his party), as debt default typically spells political instability. For example Borensztein and Panizza (2008) show that in half of the default episodes in their sample there was a change in the chief of the executive either in the year of the default episode or in the following year. In contrast, in most recent successful …scal adjustments, the government responsible for the adjustment survived the …rst electoral tests during or after the adjustment (see the Online Appendix). In sum, for the Greek leadership walking away from the deal was clearly a very costly option. Let us now compare it to agreeing to the bailout terms. It is clear that, had it lead to successful stabilization, the bailout deal’ value to the Greek party would s have been much larger than the alternative of immediate default. To be sure the deal imposed signi…cant austerity on Greece, but as we have seen savage austerity would have been in the cards in the case of default as well. Furthermore, a bailout would have prevented the collapse of the Greek …nancial sector. But as we have seen the Greek leadership might have anticipated a signi…cant chance that the programme would fail. By de…nition, failure of the May 2010 programme would have meant (and, retrospectively, did mean), for Greek leaders to soon face again an imminent prospect of disorderly default. In particular, given the front-loaded nature of the austerity programme, there was a risk of unravelling within 12-18 months. Could even this outcome be preferable to biting the bullet and defaulting right away? The answer is yes, as there was a clear option value of waiting for Greece as of May 2010.6 In particular, discussion assumes that Greece’ relevant outside option was just default. In fact it was often assumed at the time s that in case of deafult Greece would also leave the Eurozone. While it is of course logically possible for Greece to default while staying in the Eurozone, it is also possible that, conditional on defaulting, it would make sense for Greece to abandon the Euro, and impose capital controls (without capital controls the bene…ts of leaving the Euro for a country in Greece’ s situation would likely be very limited). It is also possible that such an exit would be forced upon Greece, as a result of the ECB’ refusal to supply liquidity to the Greek Banking system. Even so, it seems di¢ cult to imagine that Greece could have s avoided a deep depression had it defaulted on its debt. 6 For a broad discussion of option-vaue considerations in delaying radical decisions in politics see, e.g., Drazen (2000), especially chapter 10. An option-value interpretation is consistent with much of the informal language used by commentators to describe the agreement, e.g. “buying time” and “kicking the can down the road,” which were frequent cliches used around
5 This

6

Greece might have hoped for a softening in attitudes from the other players that could conceivably lead, in the future, to more lenient bail-out terms, particularly as regards the pace of austerity, interest rate charged, and time for repayment. Furthermore, obtaining 1 or 2 further years of …scal adjustment to bring down the primary de…cit would have made the catastrophe of default less severe, if/when it would happen. Individual rationality for the lenders Most of the other participants to the May 2010 agreement – and certainly the core, …nancially solid ones – could have scuppered the deal by refusing to subscribe to it. For reasons already discussed this would have triggered a messy, unilateral default. Hence, once again for these parties the value of walking away is the realized value of their objective function in the case of Greek default. The consequences of Greek default for other European countries were complex. There was some exposure to Greece, especially within the German and French banking sectors. Nevertheless, given the small absolute size of the Greek government bond market, few of such exposures were large enough to threaten the solvency of individual …nancial institutions, and even in these cases the respective governments could have underwritten the risks relatively easily. For example in June 2010, Germany’ (France’ foreign s s) claims vis-a-vis Greece were equal to 0.3% (0.55%) of total assets of the monetary and …nancial institutions, and the exposure to Greek government debt was 0.22% (0.17%).7 However there was a high perceived probability that a Greek default would trigger “runs” on other Euro-area sovereigns. In particular, it was thought at the time that a Greek default may shut Portugal, Ireland and, possibly, Spain out of the …nancial markets, leading to defaults in these countries as well. A Greek default was often described as the sovereign equivalent of the Lehman Brothers’bankruptcy in …nance. Such contagion would be very costly, for the countries engulfed in it, obviously, but also for other Euro area members. Exposure of core-country …nancial institutions to assets from potential victims of contagion was much more signi…cant than to Greece only, and the corresponding costs of recapitalizing domestic …nancial institutions correspondingly much larger.8 Such potentially large bailouts of domestic banks would not only be costly for the economy, but also for the electoral prospects of the heads of
May 2010 (and more generally throughout the Euro-area sovereign-debt crisis). 7 Here and elsewhere in the paper data on bank exposures are from the Bank for International Settlements (BIS). We don’ have data at the level of individual banks, so we can’ rule out that these holdings would be very concentrated in a few t t banks – particularly politically-sensitive public-sector ones. Still even in that case the losses in case of default should have been quite manageable. 8 Germany’ contribution to the …rst Greek bailout was $29.3bn. Germany’ banks at the time owned only $23bn of Greek s s government bonds. But their combined holdings of Greek, Portuguese, Irish and Spanish government debt were valued at $60bn and their holdings of debt issued by these countries’banks were worth $151bn.

7

government. In 2010 there was still much lingering resentment for the dislocations caused by the 20072009 …nancial crisis, a resentment exacerbated by government bailouts of some banks. Further direct bank bailouts would be deeply unpopular, and it may plausibly have appeared to be politically less costly to disguise such bailouts as bailouts to a fellow Euro area government.9 Aside from the potential costs of contagion and consequent bank bailouts, a Greek default may have been perceived by some heads of government as a threat to the integrity and long-term viability of the common currency. As already mentioned (footnote 5) it was often assumed that Greek default would be accompanied by Greek exit from the Euro. Some heads of government might have felt that such an exit would be seen as a precedent, to be used by other countries in the future to engineer more accommodating monetary conditions. To the extent that some of these same heads of government are committed to the European uni…cation agenda, they may also have perceived such potential threats to the Euro as very costly.10 A …nal consideration that would probably have weighed strongly against allowing Greece to default was …erce pressure from the ECB to avoid such an outcome. The ECB would have been particularly alive to the possibility of Euro-zone shrinkage or breakup, given that its very in‡ uence and existence depends on the size and existence of the Euro zone. Furthermore it was in the ECB’ political interest to make s sure that the crisis continued to be perceived as merely a sovereign debt crisis, while keeping the banking implications below the surface of the public debate. As we have seen a default would have exposed the banking implications. As preeminent monetary authority in the Euro area, such exposure would have been embarrassing and would have had the potential of weakening the independence of the Bank. Finally, the bank itself was exposed towards Greece, as it accepted Greek sovereign bonds as collateral for its liquidity provision operations. Given the enormous in‡ uence of ECB policy on Euro area macroeconomic conditions, it is very likely that the ECB’ lobbying will be highly in‡ s uential with Euro area heads of governments, so its opposition to a default is likely to have weighed in their calculations. In sum, immediate Greek default must clearly have appeared as very costly in expectation to coreaddition to the concern that a Greek default would engender fears of similar defaults by other Eurozone sovereigns, there was also a worry that …nancial markets wuold seize up due to the triggering of Greek CDS contracts. The gross and net exposures to Greek CDS and derivative contracts were about Eur 75bn and Eur 8bn, respectively. Just as in the Lehman case, there was no information on the institutions exposed to these Greek contracts, so counterparty risks would have spiked after a triggering of the CDS contract, potentially causing grave harm to the …nancial system. 1 0 We do not want to overemphasize this point, as we suspect a majority of European heads of government (and certainly their electorates) would actually be quite happy to get rid of the Euro at this stage, if they only knew how. However, it must be said that several such governments explained the decision to bail out Greece as motivated by their desire to <<save the Euro.>> To be sure, the main reason for couching the decision in these terms was that <<save the Euro>> sounded better, at the time, than <<save the banks.>> Still, we can’ rule out that some heads of government might have meant it. t
9 In

8

country counterparts. What about the bailout? A much-emphasized cost of bailouts is moral hazard. In the present case, a bailout may encourage Greece to over-borrow in the future. More importantly, a bailout of Greece may lull other Euro-area governments into believing they will also be bailed out, thus inducing them to over-borrow, too. We should note that this moral-hazard argument is less straightforward than it seems. In particular, to avoid moral hazard, core countries’ governments would have had not only to let Greece default, but also punish their own banking sectors following the capital losses on Greek (and Portuguese, and Irish, etc.) debt. Without this second ingredient banks in the Euro area would simply have learned that a sovereign default in the Euro area can happen, but is of little consequence for their own practices, as any losses will be repaid by taxpayers in their own country. Thus reassured, these banks would have continued to lend to other Euro-area governments at relatively favorable terms, which would have done little to remove the over-borrowing problem. To prevent this outcome would have required some combination of nationalization of the banks, replacement of the incumbent management, and wiping out of the existing equity holders. Appetite for such actions has been somewhat muted in continental Europe, despite a few high-pro…le cases. Even assuming that allowing Greece to default is an e¤ective means to remove moral hazard, it does not follow that any bailout deal generates moral hazard. In particular, any bailout deal is a “burden sharing” agreement, whereby the funding shortfall is …lled with a combination of austerity and …nancial assistance. It seems highly plausible that there is a range of austerity/assistance combinations that make the deal “painful enough”for the recipient country as not to wish to go through the same experience again. And, by observing the painful sacri…ces undertaken by the recipient of the bailout, other countries as well should be able to infer that over-borrowing is not rewarded. A more compelling argument against bailing out Greece is the risk of losses on the taxpayers’money invested in the bailout, in case of programme failure. One di¤erence between default in 2010 and at a later date is that default in 2010 is default on private lenders, while default at later dates might end up having a component of default on core-country taxpayers. It is true that, as we have argued, even a 2010 default on private creditors would likely have implications for taxpayers, via private-creditor bailout. Nevertheless, a direct default on a bilateral loan would likely be embarrassing politically, particularly as the loans themselves were accompanied by statements of the utmost con…dence by those extending them.11
1 1 Admittedly, bilateral inter-governmental loans are (perveived to be) senior to those underwritten by the private sector. Indeed, the March 2012 haircut on outstanding Greek bonds was entirely shouldered by the private sector. Nevertheless, as

9

On the other hand, relative to immediate default, there were option-value advantages to a one-year delay. First, some participants may have felt that a new round of negotiations in 2011 might lead to a more realistic stabilization plan for Greece. In such a scenario, default delayed could conceivably turn into default avoided. Second, and perhaps more importantly, delaying default might allow for possible changes in the cost of default itself. Delaying default would give core-country banking sectors time to reduce their own exposure to Greece. In addition, with luck other peripheral governments would get out of the sphere of contagion in the intervening period. Third, there might have been a potential blame-shifting value to postpone default. Giving Greece a di¢ cult task and waiting for the inevitable failure to accomplish it, and then blaming the default on Greece’ lack of discipline, might have appeared a politically astute manoeuvre.12 s In sum, immediate default was clearly (and correctly) perceived to carry unacceptable risks to …nancial stability in the Euro area, and to the monetary union itself. It was also …ercely opposed by the ECB. On the other hand, a bailout, while creating some novel risks, also brought a number of option-value bene…ts that might have justi…ed it even if the programme was felt to have little chance of success. I.C Pareto ine¢ ciency of the bailout Even if it was individually rational for all interested parties to agree to the provisions of the May 2010 bailout, was the deal also collectively rational? Is it possible to describe an alternative plan that, for all parties, would have (weakly) dominated the one that was implemented? In this section, we approach this question from the perspective of a set of benevolent social planners (one for each country) with symmetric and unbiased beliefs. We approximate unbiased beliefs by the view emerging from our reviews of commentary and market reactions to the May bailout, as well as our analysis of precedents for similar stabilization programmes. We describe a bargaining outcome as socially ine¢ cient if it is not on the Pareto frontier constructed from the objective functions of this set of social planners. Our analysis in this section leads to the conclusion that the May 2010 deal was indeed socially ine¢ cient. In the next section, we reach a similar conclusion for the July 2011 bailout. In Section III we discuss the kind of asymmetries in beliefs and/or political frictions that may have led to these socially the share of Greek debt that is privately held declines, further future haircuts will necessarily have to begin featuring “public sector involvement.” Hence, the 2010 decision did put core-country governments on a path to possible direct losses to their taxpayers. 1 2 One might also argue that political leaders in 2010 were just trying to shift losses on future leaders, but this seems implausible. The original deal was likely to fail by early 2012 at the very latest (because this is when Greece was to return to private sector borrowing), when both Merkel and Sarkozy were still expecting to be in power.

10

ine¢ cient outcomes. We consider two alternative plans: an orderly restructuring of Greece’ debt, and a slower path of …scal s consolidation coupled with more generous …nancing. We …nd it di¢ cult to endorse the former as a Pareto superior alternative, but we think that the latter may very well have dominated the plan that was agreed in May 2010. Alternative 1: Orderly Restructuring The alternative plan that was perhaps most popular among outside commentators was the idea of an “orderly restructuring”of the stock of Greek debt. Under orderly restructuring Greece (potentially assisted by EU institutions) would negotiate with representatives of private creditors an exchange of existing bonds with new bonds, resulting in some combination of: (i) longer maturities, (ii) lower interest rates, and (iii) lower NPV of the principal. Private creditors were expected to be amenable because the alternative – “disorderly”default –was worse. The precedent that was most-commonly cited in support of this solution was the Brady plan, that ended the Latin American debt crisis of the 1980s. Indeed, early failed attempts to deal with that crisis were also cited to criticize the May 2010 plan.13 There is no question that a negotiated lengthening of maturities, and a reduction in the NPV of the stock of debt, would have improved Greece’ predicament. To be sure, one must not overstate the bene…ts s of this route, or understate the complications. Recall that one of the problems facing Greece is the parlous state of its own banking sector. Since the Greek banking sector was heavily exposed to the Greek sovereign, the reductions in the NPV of their holdings of Greek bonds would likely come back as new liabilities for the government through the back door of needed support for the banks. This indeed has happened with the restructuring deals of 2012, even though a restructuring in 2010 would have required lower support, as foreign private lenders then held a much larger share of Greek debt. In practice, the maximum relief achievable through this means was equal to the share of debt held by foreign private lenders, which at the time was 59%. Furthermore, this deep restructuring would not have freed Greece from its need for o¢ cial assistance –unless one assumes, implausibly in our view –that the restructuring would have given Greece returned access to private borrowing. Without such an access, Greece would have needed o¢ cial loans to …nance, at a minimum, its still-large primary de…cit plus interest payments on the surviving share of debt. Depending on the extent to which it succeeded in lengthening the maturities of the surviving debt, it may also have
1 3 Particularly authoritative contemporary advocates of an orderly restructuring were, e.g. Buiter (2010a, 2010b) and Roubini (2010).

11

needed additional support to amortize maturing bills and bonds. Despite these caveats, and especially with the bene…t of hindsight, this deep restructuring-cum-o¢ cial assistance route would almost certainly have been the best outcome for Greece. From a Eurozone point of view, instead, whether these bene…ts of orderly restructuring are worth seeking depends in part on coming to a view on whether and why an orderly default is less susceptible to create a contagion e¤ect than a disorderly one. As noted, a key motivation for the May 2010 bailout was a concern that default would imperil the situation of other peripheral sovereigns. A de…nitive answer to this question would require a level of understanding of …nancial-markets behavior that is currently not available. However, it seems to us that the new information contained in an orderly restructuring is, to a …rst order, similar to that contained in a disorderly default. Namely, that Euro area government bonds are not riskless and indeed are subject to haircuts. Hence, it is di¢ cult for us to see how one can defend the view that disorderly default would have engulfed other peripheral borrowers in contagion, while orderly restructuring would have had no such spillovers. Indeed, as we discuss below, our own reading of the markets response to subsequent attempts to include (orderly) PSI in the solution to Euro-area debt crises is that, indeed, they have fueled considerable contagion.14 Another argument against ending the “sanctity”of the bonds issued by a Euro-area government focused on the impact of such a move on the ability of the ECB to conduct e¤ective monetary policy.15 Because Euro-area …nancial markets have failed to achieve meaningful integration, each country’ …nancial sector s continues to use the rate of return on that country’ government debt as the benchmark rate. This means s that if one country’ sovereign loses the con…dence of the markets, the entire economy is dragged into s a credit crunch. This makes the ECB fairly powerless to in‡ uence monetary conditions in that country, at least with conventional interest-rate tools.16 As a result, we observe enormous heterogeneity in the e¤ective stance of monetary policy in di¤erent countries. Those who made this argument concluded that, not only a sovereign default was fundamentally incompatible with the ability of the ECB to perform its duties, but also that it was imperative to rea¢ rm the riskless nature of government bonds once and for all. In sum, we think that an attempt to come up with an orderly restructuring in May 2010 would have contemporary expositions of the view that restructuring, even if orderly, could lead to contagion e¤ects see, e.g., Bini Smaghi (2010, 2011), Cottarelli et al. (2010), and Orphanides (2011a). 1 5 An early very clear exposition of this argument is in Orphanides (2011b). 1 6 There are some unconventional tools that can be used, such as the ECB’ Long-Term Re…nancing Operations of late s 2011 and early 2012. Paradoxically, and uniquely, precisely because monetary conditions were deemed quite accomodating in other economies, these unconventional policies were resorted to before the policy rate hit the zero lower bound.
1 4 For

12

improved Greece’ own position, but would plausibly have been deemed likely to weaken the prospects of s other peripheral governments. In addition, it would have further deteriorated the ability of the ECB to ful…ll its duties. It is thus far from obvious that this proposed alternative would have Pareto dominated the plan that was agreed. Alternative 2: More Generous Terms A potential set of alternatives would see Greece committing to a slower path of de…cit reduction and reform, though of course with the same …nal goal of debt stabilization. Furthermore, expected return to private borrowing would be set at a later date than envisaged in the May plan. Needless to say both these features would increase the …nancing gap, so Eurozone partners would need to make larger …nancial commitments. The attraction of more generous terms for Greece is self-evident. They would have allowed to spread the pain of adjustment over a longer time span, providing consumption-smoothing bene…ts. In addition, they might also have shifted some of the adjustment to future periods with more favorable underlying macroeconomic conditions. If thus rendered less painful, the adjustment programme may have been less di¢ cult to sell politically, with attendant improvement in the likelihood that it would be carried through. Greece’ increased chances of success would obviously have very desirable spillovers on the rest of s the Euro area. And they could easily be achieved without violating the “painful enough” requirement, so it is implausible that a somewhat more realistic plan would have created moral hazard. If anything, having been given a more realistic task might have strengthened Greece’ resolve to actually implement s it. There may well be a “conditionality La¤er curve,” which implies that austerity requirements that are too demanding end up inducing less …scal e¤ort. The obvious downside to a plan involving more generous terms for Greece is that it put more corecountry taxpayer money at risk. However, this objection loses some force when considering that payments to Greece are made in tranches, and can always be interrupted, citing failures of the Greek government to regain …scal discipline. Indeed creditor countries have routinely delayed the disbursement of tranches. This option of “pulling the plug,” common to all conditionality programmes, would limit the exposure of o¢ cial lenders. Furthermore, the expected value of losses faced by lenders are not necessarily increasing in the amount lent. To the extent that a more gradual programme has greater chances of success, lending more may lead to smaller losses - another facet of the “La¤er curve” argument made above. Some may think that more generous terms to Greece would have run into legal constraints, both at the level of the EU and at the level of individual core countries. We present a detailed discussion of the 13

role of treaty provisions and German constitutional rules in shaping the May 2010 deal in the Online Appendix.17 For the present discussion, it is su¢ cient to point out that the relevant questions of legality, if any, concerned whether bilateral loans among Eurozone members were in violation of EC treaties and/or national constitutional law - and not the size of such loans. Hence, if it was legal to lend £ 22bn, it would certainly be legal to lend twice or thrice as much. An example So far the discussion in this section has been vague about the magnitude of a more generous programme that would allow for a more gradual adjustment. To …x ideas, we now brie‡ consider a speci…c example. y This example corresponds to a “minimalist” alternative plan that calls on Greece to perform the same adjustments as in the May 2010 plan, but delays the return to private borrowing to 2014, rather than 2012. To achieve this, o¢ cial creditors would have had to cover Greece’ funding needs of EUR 68bn up s to the end of 2014 (as estimated in the May 2010 IMF review), and bond redemptions up to the end of 2013 (EUR 153bn). Greece would then be assumed to roll-over its debt maturing in 2014 (EUR 69bn). This would have required a headline bailout …gure twice as large as the one that was granted, i.e. in the order of e220bn, instead of e110bn. Needless to say this is the amount that o¢ cial creditors ended up committing after the July 2011 plan, as we will see below. In other words our minimalist plan commits o¢ cial lenders to the same overall amount as the combined May 2010/July 2011 agreements. This plan is “minimalist”as it does not allow Greece extra time to e¤ect its adjustment. But it is still bene…cial to the chances of programme success because it makes the programme less hostage to a change in market sentiment as early as 2012. As we will see below, perhaps the chief reason for the failure of the …rst bailout was that by the spring of 2011 it had become clear that Greece would not be able to return to private borrowing as originally envisaged. This triggered a hugely distracting and time consuming round of renegotiations with o¢ cial creditors, which slowed down the pace of reform and soured the political mood both in Greece and in core countries. Without this costly distraction it is conceivable that the Greek government may have managed to keep the momentum for its adjustment and reform programme which, as we will see, had been largely on track up to the time where the impossibility of returning to the markets became evident. II. The Second Bailout II.A Road to the Second Bailout main conclusion is that such legal constraints were largely not binding for the substance of the May 2010 agreement - though they may have played a more signi…cant role in other aspects of the broader strategy to deal with the Eurozone crisis.
1 7 Our

14

Despite the early pessimism, things initially went as planned in the May 2010 agreement, and for the rest of 2010 it looked like Greece might succeed in eventually stabilizing its debt-GDP ratio without a default. A large number of …scal provisions and other reforms were implemented by the Greek government, and the IMF and the European Commission issued a sequence of favorable reports on the implementation of the deal. In early 2011 Greece was still judged to be substantially on track to meet its de…cit targets. Several bailout tranches were duly paid out. Remarkably, Greece was able to issue EUR 5.7bn of bonds with maturity up to 3-years in the period October 2010-March 2011. The more positive outlook precipitated a steady decline in Greece’ bond yields until mid-October. s Unfortunately, such optimism begun evaporating in the …rst months of 2011. Three factors contributed most to the growing realization that the May plan would have to be revisited. First, Eurostat published a revised estimate of the 2009 de…cit which placed it at 16% of GDP, or 2% higher than it had been thought to have been in May. To appreciate the full impact of this discovery it is important to note that the bailout agreement expresses targets in terms of the level of the de…cit/GDP ratio, not the change. Hence, a …scal contraction in, say, 2010, deemed to be su¢ cient to reach targets if the initial de…cit level is 14%, is no longer su¢ cient if the starting point turns out to have been 16%.18 Second, and most pernicious, yields across the Eurozone increased sharply on the infamous “Dauville announcement,”by Chancellor Merkel and President Sarkozy, who said that crises after 2013, would involve “necessary arrangements for an adequate participation of private creditors.” This intent was perceived to be enshrined in the language with which the EC created the European Stability Mechanism (ESM) in March 2011, resulting in persistently high spreads over bunds for all peripheral countries, and foremost for Greece. Third, beginning in March serious slippages in Greece’ implementation began to appear. This severe s slackening likely re‡ ected a variety of reasons. To some extent it may have been part of a “chicken game” vis-a-vis the Eurozone partners, to put pressure on them to come to a quicker resolution of the new bargaining round. It could also be that, while there was uncertainty on whether a new bailout (which was essential to avoid default) was forthcoming, the political cost of implementing reforms that may
1 8 There is an interesting discussion to be had on the rationale for couching bailout-agreements in terms of de…cit-GDP targets rather than in terms of changes to the absolute value of the de…cit. Targeting levels rather than changes means that the recipient of the bailout carries all the risk from data revisions. Targeting the ratio of the de…cit to GDP rather than the absolute level of the de…cit means that the recipient of the bailout carries all the risk from ‡uctuations in the growth rate. In IMF-led bailouts such asymmetric allocation of risks may be of relatively little consequence as the IMF has the ‡exibility to renegotiate relatively easily and frequently. But for a bailout involving a large number of creditors renegotiation is immensely costly and slow, as the events we are about to recount demonstrate, so the recipient of the bailout does e¤ectively carry a disproportionate share of the risks.

15

later appear to have been pointless was perceived as higher. In addition, the program for 2011-2013 was emphasizing structural reforms and privatization, which are politically harder to implement than …scal contraction. These factors implied that Greece was no longer meeting its reform and stabilization targets and that it was not going to regain access to private funding by the time envisaged in May 2010. In other words, the May plan was now “o¢ cially” outside the feasible set. The new round of bargaining took extremely long, and only came to fruition in July 2011, some six months after it had become clear that the original plan would not work. We conjecture that there were two (related) reasons for the long delay in coming to a new agreement. First, the surplus from avoiding default must have been felt to have shrunk. Clearly any agreement would have had to have a combination of more aggressive …scal tightening for Greece, and additional …nancial assistance from the rest of the Eurozone, making an agreement correspondingly costlier. On the other hand, the costs of a default had arguably not changed very much, so the surplus was smaller. Second, with a smaller surplus on the table, participants naturally looked for additional margins of adjustment. Germany in particular begun arguing for forms of reduction in the value of debt held by private creditors.19 This attracted …erce opposition from the ECB, and the bargaining became all the more complicated and sluggish. Eventually the agreement of July 2011 featured concessions on all three sides: Greece agreed to an even tougher austerity and reform programme, including the disposal of large numbers of state-owned assets; countries in the Euro zone (through the recently-created European Financial Stability Facility, or EFSF) committed to a new EUR 109bn in bailout funds, and the ECB had to acquiesce to an element of PSI. II.B Evaluating the Second Bailout Our evaluation of the July 2011 bailout is similar in many respects to the one for the May 2010 bailout. First, it seems quite clear that the actions to be undertaken were outside of the feasible set. The last column of Appendix Table A1 shows changes in the primary de…cit called for by the plan in each of its years. The new plan calls for two consecutive years of massive …scal adjustment, a feat rarely encountered in previous successful large austerity programmes. There is only one precedent of a country succeeding in implementing an average annual primary-de…cit reduction larger than the one Greece was to undertake, greater willingness to contemplate PSI to a Greek bailout may have been partially due to the exposure to Greece of German banks having fallen from $60bn to $25bn between May 2010 and July 2011 (foreign claims fell from $31bn to $21bn and exposure to Greek government bonds felll from $23bn to $12bn). It is also the case that while CDS spreads on Greece, Portugal, and Ireland had increased dramatically since late 2010, those on all the other Eurozone countries had fallen slightly. This may have led Germany to underestimate the risk of contagion.
1 9 Germany’ s

16

and none that has achieved a comparable cumulative reduction over a similar number of years. Recall that the comparison programmes are the most aggressive on record in the OECD in the last 40 years. Meanwhile, the overall macroeconomic and political backdrop had, if anything, further deteriorated, as can be seen in the last columns of Appendix Tables A2 and A3. Second, it is probably still possible to build an option-value case for the individual rationality, for each participant, of accepting the deal instead of walking away from it. This case would have many similarities with the cases made in respect to the 2010 agreement, so it will not be made here to avoid repetition. Third, from a collective standpoint the July agreement appears even more suboptimal than the May one. One reason for this is similar: as the July plan was more clearly unfeasible, the case for the superiority of a deal allowing Greece more time and a more gradual adjustment, so as to bring it on the left side of the conditionality La¤er curve, is even stronger. Another reason is novel to the July agreement: the inclusion of the PSI component. Because this is the novel element, we focus the rest of the discussion on it. The agreement committed Greece to open negotiations with representatives of private lenders to achieve a voluntary reduction in the value of the debt of approximately 20%.20 Needless to say, by specifying a target haircut of 20%, the agreement e¤ectively made it an upper bound on the size of the equilibrium outcome. With a publicly stated goal of 20%, the Greek negotiators could hardly have picked an initial negotiating position of, say, 50%. Hence, it was built into the bailout agreement that the haircut on private holders was to be between 0 and 20%. As private creditors (ECB excluded) held an estimated 58% of the overall stock of Greek debt outstanding, the absolute upper bound on the debt relief that might come from the deal was a 12% reduction of the debt/GDP ratio (20%x58%). The mid-point of the bargaining range would have delivered a 6% reduction. In other words, the PSI component of the deal was little more than symbolic, and provided no meaningful debt relief. Despite the symbolic impact of PSI on the NPV of Greek debt, some commentators saw some bene…ts in the plan because private creditors were expected to swap existing claims on Greece with ones bearing a longer maturity. As a consequence, a certain amount of bonds originally due to mature before 2011 and 2014 would now mature at later dates. This lengthening was argued to be bene…cial to o¢ cial creditors, because, under expectations prevailing in July 2011, it would have been di¢ cult for Greece to persuade actual provisions for PSI in the actual Statement issued by the EC at the end of the meeting are quite ambiguous. Point 5 refers to a <<net contribution of the private sector .... estimated at 37 billion euro>>. However footnote 1 also says <<For the period 2011-2019, the total net contribution of the private sector is estimated at 106 billion euro.>> The 20% …gure emerged from subsequent commentary and debrie…ngs. Since PSI had to be negotiated with private creditors anyway, speci…c …gures in the Statment were little more than opening bids in the negotiations. For more details see the Appendix, as well as Zettelmeyer, Trebesch, and Gulati (2012).
2 0 The

17

private creditors to rollover expiring bond issues (at acceptable rates) over this period. Hence, in the absence of the bond swap, o¢ cial creditors would have had to commit even more funds to the July bailout, in order to provide for the amortization of maturing medium and long term debt held by the private sector. According to this view keeping this debt in private hands was a way to retain the option value of bailing in private creditors in the future. As we discuss in greater detail in the appendix, there are two sets of reasons why these arguments don’ bear scrutiny. First, the debt swap had to be voluntary, and there was immediate considerable t skepticism on the extent of take up. Second, and most importantly, the debt swap had a number of adverse consequences which made its net impact on Greece’ …nancing needs between 2011 and 2014 much s smaller than its gross impact suggested - if not indeed negative. These costs included the necessity of recapitalizing Greek banks for their NPV losses on their huge holding of Greek government bonds, the costs of …nancing a “credit enhancement programme” in which the new bonds issued by Greece would be guaranteed by AAA-rated bonds issued by the EFSF, and several others. Indeed the swap ‡ oundered in the end for lack of clarity on the terms, lack of interest, and dawning realizations that the costs the swap entailed dwarfed the bene…ts in terms of longer maturities. On the other hand, there are clear indications that the presence of a PSI component in the July agreement increased the costs and the complication of the broader European debt crisis. It seems likely, for example, to have been a signi…cant factor in dragging Italy into the crisis in the late Spring and Summer of 2011. As we show in the Online Appendix, the yield on 5-year Italian bonds, which had been quite stable over the months of May and June, rose dramatically in the run-up to the EC council. As we also document, the early part of July saw an acceleration in the number of o¢ cials publicly predicting that the July EC meeting would agree to include some form of PSI. Hence, a possible interpretation of this spike in Italian spreads is that it re‡ ects pricing-in of the contagion from Greek PSI.21 ;22 With the third biggest economy of the Eurozone engulfed, the extent of the existential threat to the Euro project rose to a whole new level. Since PSI provided no meaningful bene…ts in terms of debt reduction, and plausibly caused considerable havoc through contagion, it seems that not having included PSI in the …nal deal might have been Pareto improving. EC President Herman van Rompuy summed this be sure the hypothesis of a PSI component had been on the table since well before July - early June at the latest. But it was only in early July that it became clear that PSI was more likley than no PSI. 2 2 Admittedly yields fell for a day or two upon the bailout announcement. However throughout the crisis it has been common for spreads to decline for a couple of days after a major EC council, possibly in response to EU leaders’triumphal announcements that the crisis has been comprehensively dealt with for good. Reality apparently sinks in with a few days’ lag.
2 1 To

18

up on December 9, 2011, when he said “As regards the Private Sector Involvement (PSI), we have made a major change to our doctrine: from now on we will strictly adhere to the IMF principles and practices. Or to put it more bluntly: our …rst approach to PSI, which had a very negative e¤ ect on the debt markets,
23 is now o¢ cially over.” ;24

III. Understanding Suboptimal Deals in the EU We have argued that the May 2010 and July 2011 decisions were socially ine¢ cient, in the sense that a set of benevolent planners (one for each country), all sharing the beliefs expressed by the vast majority of commentators and analysts at the time, would have been made better o¤ by deals allowing more time to Greece, and (in the case of the 2011 bailout) no PSI component. It is conceivable that some of the key political players held more optimistic beliefs than the experts and the other parties to the deal. When bargaining parties hold biased or heterogeneous beliefs, the relevant notion of a Pareto frontier takes these beliefs into account. On the “perceived” Pareto frontier no party can be made to perceive himself to be made better o¤ (based on his beliefs) without having another party perceive herself to be made worse o¤ (based on her beliefs). Parties may well succeed in bargaining to the perceived Pareto frontier, but clearly they will not typically seem to have reached the Pareto frontier as de…ned by a common and unbiased set of beliefs. If, for example, Chancellor Merkel felt that it was economically and politically feasible for Greece to implement the May 2010 programme, we would have a rather simple explanation for the social suboptimality of the observed outcome. While we don’ categorically rule this possibility out for the May 2010 agreement, we are skeptical about t its applicability to the July 2011 one. At that time, the unrealistic nature of the paths laid out for Greece had to be apparent to all. Furthermore, given the widespread public commentary on PSI in the months and weeks before the July deal, it seems implausible that the German party did not understand that the symbolic PSI provided no material bene…ts in terms of alleviating Greece’ funding needs (and hence s o¢ cial creditors involvement). Hence, as long as they put a strictly positive estimate on the probability of contagion from PSI (however small) they should have preferred no PSI to symbolic PSI. We therefore turn to discuss possible political frictions that could have contributed to moving the bailout agreements inside the (social) Pareto frontier.
2 3 http://www.consilium.europa.eu/uedocs/cms_Data/docs/pressdata/en/ec/126657.pdf

Emphasis added. be clear, we are not saying that zero PSI would have dominated any amount of PSI. A meaningful PSI component which would have truly reduced the present value of the debt could concievably have made at least some participants better o¤ than no PSI at all. We are simply saying that no-PSI dominates purely symbolic PSI.
2 4 To

19

II.A Voter intransigence The simplest political friction capable of pushing bargaining among EC members inside the Pareto frontier is a version of the asymmetric-beliefs story, but applied to voters rather than to leaders. Consider the 2010 bailout. Perhaps core-country voters then underestimated the severity of Greece’ …nancial and s economic position, and overestimated the chances of success of the tough programme that was agreed. More generous terms would have been perceived as putting more of their money at risk than necessary, and perhaps also providing bad incentives to Greece and other peripherals. These perceptions being too entrenched for political leaders to modify through persuasion, they ended up constraining the set of politically-feasible bailout packages. In this view, then, in May 2010 and July 2011 core-country leaders did “the best they could do” given political pressure at home to strike a tough bargain. They knew that they would have to do it all over again in about one year, but they hoped that, at that junction, German voters would have been “educated” by the failure of the previous years’ programme, and would thus be willing to accept an expansion in commitments that they would have …ercely resisted on the previous round. One could argue that, if this was the signatories’thinking, the plan did not quite work. Under the plan outlined above, German voters’attitudes to Greece should have softened between the …rst and the second bailout, as voters learned of the true plight of the Greek economy and the harshness of the consequences of the programme on the Greek people. Instead, attitudes towards Greece seem to have hardened. On April 28, 2010 (eve of the …rst bailout) 49% of respondents said “yes” to the question “Should Greece be excluded from the Eurozone?,”while on June 18, 2011 (just before the second) the “yes”answer had risen to 58%!25 This is an ex-post outcome, however, and does not rule out the possibility that political leaders may have hoped ex-ante for a di¤erent evolution of voters’attitudes. A somewhat stronger objection to the argument that voter intransigence was a binding constraint is that Chancellor Merkel was not up for re-election until September 2013. It may be that more generous terms in May 2010 would have damaged her short-run popularity, but her goal should have been to maximize support three years later. Arguably, this goal would have been better served by presiding over a single successful bailout (however unpopular it may have been at the time), rather than a series of
2 5 In the one intervening poll in January 2011 the percent yes was down to 40%. This could re‡ect the good start Greece had had on the …rst programme, or the fact that the January poll had been conducted by Allenbach, while those in April 2010 and June 2011 were by Infratest-Dimap, raising the possibility of di¤erences in polling methods (Allenbach tends to have much higher proportions of “don’ know”s). There are many other German opinion polls regarding Greece, and the t Euro crisis but very few questions are consistently repeated with the same wording over a time span covering both bailouts. The one on Greece’ exclusion seems to be the only exception. s

20

botched ones, particularly as the latter path involved cumulative commitments to Greece (as of 2013) at least as large as under the former, lower overall chances of programme success as of 2013, and all the reputational costs implied by having repeatedly subscribed to doomed deals between 2010 and the time of the election. To this, however, one could respond that the Chancellor was concerned with the outcome of state elections (7 in 2011, 3 in 2012), and in particular that her party (and her coalition partner) might be punished at the state level for bailout terms seen as too generous. It is di¢ cult to rule out with con…dence that concern with short-term state-level outcomes would be su¢ cient to overcome the longer-term considerations applying to the 2013 federal elections. In the end, perhaps the most important question regarding the voter-intransigence interpretation is whether it is plausible that the political cost of bailing out another country would be very sensitive to the size of the bailout. Voters, even sophisticated ones, become fairly insensitive to …gures that are orders of magnitude outside their practical experience. Some voters are clearly hostile to the idea of bailing out Greece. But conditional on a bailout taking place, one may doubt that they will scrutinize the exact amount carefully. Had a core-country head of government come out of the bargaining room with an announcement that she had committed £ 44bn rather than £ 22bn, would she really have borne a much larger loss in popularity? II.B Communication Frictions In order to discuss the second political friction that could explain suboptimal EC bargaining outcomes we focus, for concreteness, on the symbolic PSI component of the 2011 agreement. On this dimension, we believe that the “objective”payo¤ functions of the key players can plausibly be assumed to take the forms shown in Figure 1. On the horizontal axis we measure the size of the haircut to be negotiated with the private creditors. On the vertical axis we measure the loss functions of a typical core country (henceforth Germany, for brevity) and of the ECB. The ECB simply wants as little haircut as possible, so its loss function is monotonically increasing. Germany would like to get the private sector to share in the burden, so we assume that its loss function is globally minimized at some signi…cant level of the haircut, denoted “Germany’ bliss point” in the …gure. Germany’ loss monotonically increases as the size of the haircut s s diminishes. However, our discussion above implies a discontinuity at the origin. Because of the negative e¤ects the symbolic PSI provisions had on other peripheral countries, a zero haircut is discretely to be preferred to a purely symbolic but strictly positive haircut. In other words, no PSI whatever is a local minimum in Germany’ loss function - and more generally Germany’ preferences are not single peaked. s s In the Figure we have also denoted by “Bargaining outcome”the empirically observed outcome, of a very 21

small, symbolic haircut. Figure 1 Here The mechanism we wish to discuss is based on a plausible friction in communication between politicians and voters. Some background on context is in order before describing our hypothesis. EU summits, at least in crisis periods, are focal political events that are eagerly anticipated. In the weeks leading to each summit there is pervasive coverage and extensive public commentary on the possible and/or desirable outcomes. In such a climate, heads of governments and their spokespersons are under intense pressure to comment on their goals and strategies. Some of them are even required to report to Parliament on their negotiating position in the upcoming meeting. Needless to say, such public pre-summit positioning has important implications for the head of government’ negotiating stance. It implies that her performance s will be judged by voters on the distance between the bargaining outcome and her stated bargaining goal, giving her powerful incentives to negotiate hard for her stated position. Indeed, it is very likely that some leaders use such public pronouncements simultaneously as a commitment device and a signalling tool to other leaders. By publicly stating their negotiating position they make it costlier for themselves to fail to achieve an outcome close to it. Not only this strengthens their own resolve but also increases their bargaining power, as other governments now know that for this head of government the cost of scuppering a deal relative to accepting one that is distant from her preferred option are relatively small. Imagine now that there is some communication friction which implies that political leaders can only convey fairly simple messages to voters concerning their negotiating strategy. This does not have to be due to voter irrationality or lack of sophistication, though these are undoubtedly plausible assumptions in their own right. It could be that voters have rational inattention towards politics, hence devoting only limited scarce cognitive resources to it. Or it could be that voters rationally distrust excessively complicated messages as they believe that they could be used by politicians to inject noise in their assessment of the politician’ performance.26 s Under such communication and/or cognitive frictions it is likely that heads of government will tend to limit the amount of information contained in their pre-summit statements. In particular, a plausible working hypothesis that also seems consistent with casual observation is that such statements will be
2 6 There is a literature on communication frictions between politicians and voters [e.g. Cukierman and Tommasi (1998a, 1998b)] but we are not aware of models in which these take the form of constraints on the form of the messages that can be sent. However, in the “cheap talk” literature there is a tradition of formulating models where senders are limited to “coarse” messages, for example announce a discrete interval rather than a speci…c real number, and some of these tools may be relevant to model the mechanism we are sketching [e.g. Crawford and Sobel (1982), Austen-Smith (1990, 1993)].

22

limited to descriptions of the head of government’ ideal point, and not of the entire pro…le of her loss s function. Similarly, it is very likely that voters will automatically assume that the objective function is single peaked, and will judge the outcome strictly in terms of its distance from the stated ideal point. In most cases single peakedness is probably a good assumption, so this heuristic is a reasonably e¢ cient way to deal with communication frictions between politicians and voters. However it is immediately apparent that when single-peakedness is violated constraints on communication can become quite costly. In particular, knowing that she will be judged by voters on the distance of the negotiating outcome from the ideal point, the head of government who is (su¢ ciently) concerned with her electoral prospects might act to minimize that distance. With a non single-peaked loss function this is not the same as minimizing the loss.27 In the Online Appendix we provide a broad overview of public statements by European political leaders in the months and weeks leading up to the July 2011 agreement. Consistent with the discussion above, we …nd that Germany was deeply committed to PSI going into the negotiations, suggesting that a deal not including some amount of PSI would be politically quite costly to the Chancellor. For example, speaking on July 17th, Chancellor Merkel was in no doubt that private investors would share some of the burden of the new deal for Greece: “The more we can involve private creditors now on a voluntary basis, the less likely it is that we will have to take next steps, ... But most important is that Greece does its homework and private creditors have to be involved.” On the other hand, throughout the negotiations, the ECB maintained its …rm resistance to a credit event of any form, instead insisting greater government …nancing should be provided. This was clearly expressed during an early July press conference by Trichet with the remarks, “no credit event, no selective default, no default. That is the message of the Governing
28 Council.” In an interview with FT Deutschland on July 14th, presumably directed at a German audience,

the President of the ECB explains further: “If a country defaults, we will no longer be able to accept its defaulted government bonds as normal eligible collateral.” “The governments would then have to step in themselves to put things right.” He also criticized politicians’ lack of “verbal discipline” in reference to their mixed signals in public.29
2 7 To some extent this mechanism is also vulnerable to the objection that elections were not due until 2013. But the critique does not apply as strongly as – in the game we described – the voters are not only learning about the fundamentals of the Euro crisis. They are also making inferences about their head-of-government’ type, particularly as regard her toughness s and e¤ectiveness in multi-lateral negotiations. Because high-stakes multilateral bargaining sessions are infrequent, there are relatively few such learning opportunities, and “success” at one such meeting could have persistent e¤ects on a leaders’ perception by the voters. 2 8 See transcript of July 7 2011 press conference. 2 9 See Reuters article, Trichet: ECB would reject Greek bonds as collateral .

23

II.C Bargaining Frictions A third possible explanation for suboptimal bargaining outcomes in the EU is based on the dynamics of bargaining with a time limit. To motivate this conjecture it is important to recall that ECOFIN and EC meetings, where key decisions are ironed out, are infrequent and time-limited a¤airs. A typical meeting will begin with a meal and go on until an agreement is reached or the meeting is broken up. If the matter is relatively uncontentious and the key points have already been agreed in advance by the diplomatic sta¤ the meeting is simple and brief, and mostly an opportunity for communication with the voters. However on di¢ cult and controversial matters the meeting is a true “end game” negotiating session. Ample anecdotal evidence indicates that as the negotiations continue over many hours tiredness and personal animosities among the heads of government become meaningful factors in the negotiation process. Similarly, anecdotally it appears that negotiations are sometimes concluded not so much out of a feeling of having achieved the best possible deal but out of sheer exhaustion and/or impossibility to overcome hardened attitudes. It should be possible to conceptualize the bargaining process in a way that makes some sense of these anecdotes. Consider the following bargaining protocol. At the beginning of the bargaining session a “political auctioneer”asks all participants to state their most preferred outcome. For Germany this will be, say, a 70% haircut on private creditors, while for the ECB it may be no haircut at all. The political auctioneer then tables a proposal that is a convex combination of each participant’ most preferred outcome, with s weights equal to each participant’ bargaining weight (assumed to be known to the auctioneer). Hence, s for example, if the ECB has a large bargaining weight on the issue of PSI, the auctioneer will table a proposed haircut that is between 0 and 70%, but closer to 0 than to 70%. Once again, in most cases it is likely that participants have single-peaked preferences, and under single-peaked preferences this bargaining protocol generates a Pareto e¢ cient proposal. It is therefore plausible that negotiations will follow this protocol in general, and will often quickly converge to outcomes on the Pareto frontier. However occasionally some parties to the negotiation will have non single-peaked preferences and in these cases clearly the proposal of the auctioneer is ine¢ cient. It may be that this ine¢ ciency will be felt by the participants but it is also likely that this will lead to a time-consuming search for a better deal. Furthermore it seems likely that the auctioneer will for a considerable time continue to search in the interior of the set delimited by the participants’ideal positions, perhaps because it attributes the failure of the …rst proposal to private information on the value of outside options and hence bargaining power. To be sure, should the search process be allowed to continue inde…nitely it is 24

likely it will lead to the discovery of a Pareto e¢ cient deal. But, as discussed above, the bargaining session is time limited, as exhaustion, bad feelings, or important commitments back at home make it necessary to break o¤ the session. The time at which the session must end may well be a random variable, but this does not change the reasoning. When the time to end the session comes, the entire negotiation boils down to a “take it or walk away”decision by each participant, where “it”is the latest iteration in the search for a deal. If the end time comes too early, the deal on the table will still be inside the Pareto frontier. As we discussed, “take it” can dominate “walk away” for everyone even if “it” is not on the Pareto frontier.30 Clearly this mechanism presumes an element of asymmetric information, in that the political auctioneer that generates proposals must have noisy information on the shape of the various parties’loss functions. Otherwise it would be able to identify the Pareto frontier. Limits on information on the part of the auctioneer can persist because participants have an incentive to hold out for a better deal. For example, Germany prefers no PSI to symbolic PSI, but does not wish to reveal this because it hopes the auctioneer’ s next iteration will be “to the right” of the current proposal. Hence, the mechanism combines elements from the literature on deliberation [e.g. Austen-Smith and Feldersen (2005, 2006), Coughlan (2000), Eliaz, Ray, and Razin (2007), Gerardi and Yariv (2007, 2008), Meirowitz (2006), Persico (2004), and Lizzeri and Yariv (2011)] with elements from the literature on wars of attrition [e.g. Alesina and Drazen (1991), Drazen and Grilli (1993), Ponsati and Sakovics (1996), Hsieh (1997)]. The former literature is relevant for its emphasis on how protocols for committee decision-making a¤ect decision outcomes, while the second literature shows how asymmetric information leads to delays in bargaining.31 Notice that the mechanism outlined above encompasses the possibility of a role for asymmetric beliefs among the participants. As already implicit in our discussion so far, there was sharp disagreement among commentators on the gravity of the contagion risk from orderly default/PSI. It seems possible that similar disagreements existed among the parties to the July agreement. In particular, the ECB (together with most peripheral governments) was clearly extremely worried about contagion, while is seems possible that Germany felt it to be less likely. These di¤erences in beliefs could undoubtedly have contributed to shaping conjectures above are mainly inspired by anecdotal newspaper accounts (based on participants’leaks) of the dynamics of EU summits. Additionally, they may be in‡uenced by a dozen-year experience of attending faculty meetings. 3 1 In the deliberation literature the most relevant contribution is Lizzeri and Yariv (2011) which studies sequential deliberation : a jury decides every period whether to keep sampling information or stop deliberation and make a decision based on the information hitherto uncovered. But in their paper there is no time limit to the possible length of deliberation. More importantly, there is no fundamental con‡ict of interest on the …nal outcome, nor asymmetric information on the preferences of members of the jury. These features are, of course, central to war of attrition models, which o¤er endogenous mechanisms that generate delay in solving problems. Another paper that o¤ers a relevant ingredient is Eliaz, Ray, and Razin (2007), who emphasize how members of the decision-making body may decide to accepet a feasible (but for them suboptimal) outcome of the deliberation rather than having “disagreement,” i.e. a collapse of decision making equivalent to our “walk away” option.
3 0 The

25

the objective functions that Germany and its counterparts brought to the negotiating table, i.e. they made Germany broadly more “pro-PSI”and the ECB and others more “anti-PSI” However, as discussed above, . di¤erences in beliefs alone do not seem to be su¢ cient to explain the suboptimal outcome in question. IV Conclusions This paper has focused on the political-economy of Eurozone bailout agreements. It has looked at the cases of the May 2010 and July 2011 bailouts of Greece. In both cases accepting the conditions of the agreement, rather than walking away and triggering an immediate default by Greece, can plausibly be argued to have been individually rational for each of the parties. However we …nd it harder to argue that the two agreements were also collectively rational. In both cases, giving Greece more time to reduce the de…cit before returning to borrowing on private markets might have increased the chances of success of the stabilization programme, at no large increase in the political or …nancial cost borne by the lenders. Furthermore, in the case of the July 2011 agreement, omitting the symbolic PSI element would have reduced the likelihood of contagion to other Euro area peripheral countries, without materially increasing the …nancial burden on o¢ cial creditors. We have sketched some possible arguments for why EU-wide bargaining could lead to ine¢ cient outcomes. Among these, the two we found to have most potential rely on the existence of non single-peaked objective functions for at least some of the participants. In one argument this non single-peakedness interacts with the infrequent and time-constrained nature of EU negotiations (which is due to the di¢ culty of bringing together 27 heads of government and assorted heads of European institutions for more than a few hours every few months). In these negotiations, a “political auctioneer” will tend to search for deals that are convex combinations of the participants’ideal points. Under non single-peakedness and time-limited search this will tend to produce outcomes that are acceptable relative to walking away, but inside the Pareto frontier. In another argument non single-peakedness interacts with communication frictions that limit the ability of heads of state to provide accurate descriptions of the entire pro…le of their loss functions. This induces voters to judge the negotiating outcome on the basis of its distance from the global minimum of the loss, rather than based on the realized value of the loss. Electoral considerations imply that the head of government inherits this monotonicity in shaping her attitudes during the negotiation. It is useful to contrast these mechanisms with the situation that arises in “normal” bailout situations, where the emergency lender is the IMF rather than the European Council. It is not the case that the IMF’ loss function is necessarily always single peaked, though we suspect this is likely to be the case s more often. The main di¤erence is that neither the problem of infrequent and time-limited negotiations 26

nor the problem of communication with domestic electorates is particularly relevant in the case of the IMF. As already mentioned, the IMF is able to renegotiate often and at leisure, while the 27 heads of state can only meet infrequently and for a few hours. This makes it much easier for the IMF and its partners to search exhaustively for a solution on the Pareto frontier. Similarly, the IMF is not compelled to make public announcements of its “game plans” to its constituency before opening discussions, nor is it forced by communication frictions with voters to oversimplify its negotiating position. In other words, domestic political considerations are paramount in EU bailouts, but not in IMF ones. If the EU is bent on continuing to take the lead in future bailouts it should give careful considerations to ways to limit the pernicious e¤ects of these frictions.

27

Appendix Appendix 1. Assessing the Feasibility of the May 2010 Programme Did the May 2010 agreement commit its parties to feasible actions? In particular, given the severity of the austerity programme it asked Greece to implement, was the plan a realistic one, on an ex-ante basis? Any answer to this question is necessarily highly speculative, but this section takes a rough stab at it. In Table A1 we report data from the 8 largest …scal adjustments in OECD countries since the 1980s, as well as Greece’ previous stabilization in the early 1990s.32 For each episode, each column shows the percentages point change in the primary de…cit achieved in each year of the programme. It also shows the average annual and cumulative percentage change in the primary de…cit, and (implicitly) the programme duration. In the penultimate column, we present equivalent numbers for the May 2010 austerity programme. Note that while the …gures in the …rst eight columns are ex-post numbers, i.e. they re‡ adjustments that ect were actually achieved, the column for Greece reports ex-ante numbers, i.e. what was expected of Greece going forward. Table A1 Here The overall impression from Table A1 is that the austerity programme that Greece was called on to implement by the May 2010 agreement was tough but not unprecedented. The numbers for the Greek plan are of comparable magnitudes to those of these very large adjustments, and correspondingly call for a vast e¤ort. But there is no dimension of the programme that appears to be an outlier. There are countries that have had larger average annual and cumulative percentage-point declines in the primary de…cit, larger declines in individual years, more front-loaded, and longer programmes. However in order to fully evaluate the feasibility of the programme it is also important to look at the macroeconomic context in which the programme is to be implemented. Table A2 Here The real GDP growth experiences during the previous adjustments are presented in Table A2 together with the IMF forecasts for Greece’ growth during its adjustment. The majority of the previous ads
3 2 The Table shows the largest eight episodes of cumulative reduction in the primary de…cit/GDP ratio in a sample of 20 OECD countries from 1970 to 2010. We consider episodes of …scal tightening all years in which the primary de…cit/GDP falls. We cumulate the decline in the primary/de…cit over consecutive years in which it falls. Note that we would select a similar set of countries/years if we used a measure of cyclically adjusted primary de…cit/GDP ratio to select episodes of …scal tightening. Countries included in the sample are Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Japan, the Netherlands, New Zealand, Portugal, Spain, Sweden, Switzerland, United Kingdom, United States. The methodology is quite similar to the one used in Alesina and Ardagna (2009) in their study of the determinants of success in …scal consolidations.

28

justments took place with strong average annual growth in excess of 3 percent. The May 2010 forecast for Greece’ average annual growth is lower than any of the realized average growth rates in previous s adjustments. Only the Greek …scal adjustment of 1990-1994 had a somewhat comparable stagnant average annual growth rate, but it is perhaps also important to note that table A2 does not show the fall of Greece’ real GDP in 2009, the year immediately before the start of the adjustment.33 Certainly no s previous adjustment of this scale had commenced in the middle of a 3 year long recession with an 8.4% peak-to-trough contraction as had been forecast at the time. Even if the more solid growth in the later years of the adjustment had materialized, Greece’ real GDP would still have been lower in 2015 than it s had been at its 2008 peak. Table A3 Here Table A3 analyzes the timing of the adjustments in the context of respective business cycles. A similar pattern to Table A2 emerges, with the majority of the previous adjustments taking place as the output gap increases from start to …nish. These adjustments therefore took place when the economy was either in the recovery phase or the boom phase of the business cycle, which contrasts starkly to the timing of the 2010-2015 Greece adjustment which was to take place in the opposite circumstances. The forecasts for both the average annual decline and the total decline in the output gap under the May 2010 program were at least twice as great as the only other adjustment with a negative change, Greece 1990-1994. Table A4 Here The maximum levels of general government gross debt implied by the programmes are additionally reported in Table A4. The maximum levels experienced by all of the countries in the previous adjustments were to be surpassed by the 2013 forecast for Greece, which at 149.2% of GDP would become the largest of any advanced economy since 1980 - aside from Japan which has a debt market with unique characteristics. Table A5 Here Tables A5 and A6 document the current account changes, changes in real e¤ective exchange rates (REERs), and exchange rate regimes during the adjustments. These indicators are of critical importance when analyzing the program of May 2010 because being in the Eurozone meant that Greece could not devalue its currency against its most important trading partners.
3 3 At the time of the May 2010 Country Report, the 2009 GDP growth was given as -2% but this has since been revised to -3.3%. See Hellenic Statistical Authority Press Release 05-10-2011 for more details on Greece’ GDP revisions 2005-2011. s

29

Table A6 Here Table A5 shows that Greece had a current account de…cit far greater than any of the other countries at the onset of its adjustment. Table A6 shows that the listed countries had varying degrees of exchange rate ‡ exibility, from the fully ‡ oating Canadian Dollar to the pegged Danish Krone. These di¤erences are mirrored in a wide variety of outcomes for the REER during the stabilization episode, with some countries experiencing overall depreciations while others appreciated. However, a closer look at the table reveals that even those countries that experienced real appreciation during the period of adjustment, had undergone signi…cant real depreciation shortly before undertaking the stabilization programme.34 It may be important that the two least ‡ exible exchange rate regimes in the sample, Denmark and The Netherlands were the two countries that saw a signi…cant deterioration in their current account balance. Conversely, the two countries that saw the greatest improvements to their current accounts were Sweden and Finland who both ‡ oated their currencies having withdrawn from the ERM immediately before their adjustments. Figure A1 Here To try to compare these examples to the position of Greece in May 2010, Figure A1 plots cumulative CPI indices for Greece and Germany since the Euro inception, using historical data up to 2010 and May 2010 forecasts for the subsequent period. A widening of the gap between the two is a (rough) proxy for REER appreciation. It is clear that by this metric Greece’ REER (i) appreciated signi…cantly during the s pre-crisis period and (ii) was expected to go on appreciating for several years into the adjustment period. The evidence from the previous …scal adjustments contained in Tables A5 and A6 together suggests that the current account improvement forecast for Greece in May 2010 were optimistic given the fact that Greece had become uncompetitive in the decade since joining the Euro and, moreover, it was not expected to devalue its REER substantially through the adjustment. In sum, the May 2010 agreement committed Greece to an adjustment that, while not unprecedented in its size, was unprecedented in the adversity of the macroeconomic context within which it was to be implemented. Appendix 2. Bene…ts from PSI in the form of longer maturities
3 4 The exception to this rule is Greece in the early 1990s. That adjustment occurred when its in‡ation was high (double digit throughout). Despite Greece’ nominal e¤ective exchange rate depreciating by over 30% during the adjustment, the s REER based on CPI is likely to have appreciated.

30

It is important to begin by saying that there is immense confusion regarding the details of the bond swap envisaged in the July 21st agreement. The o¢ cial documents are incredibly opaque and hard to reconcile with each other; participants to the agreement gave di¤erent accounts in the subsequent press conferences; and even in the following weeks di¤erent accounts persisted in the press and in documents issued by …nancial-market participants. This of course underscores the chaotic nature of EU-Council decision-making, which is one of the key themes of our paper. But it is particularly important for the discussion at hand because it means that, quite frankly, nobody knows what core-country governments really thought they had exactly agreed to. With this caveat, we make two main points. The upper bound on the extent to which the bond swap reduced …nancing needs up to 2014 is euro 54bn. The …gure comes from a press release by the Institute of International Finance (IIF) issued to coincide with the EC meeting.35 The press release and the technical documents that accompany it do not provide any details on how the …gure had been arrived at, other than to say that it is based on a “target participation rate of 90%” But the 90% participation rate was immediately greeted with considerable . scepticism. Representative of the media queried the 90% number as early as in the press conferences of various council participants in the immediate aftermath of the meeting. Many noted that the list of banks announcing their participation missed many important players (also, some of the ones adhering to it had no Greek bonds to exchange!), and several major banks were distinctly coy about their participation in the days and weeks after July 21st. Many also noted the extreme opacity and complication of the menu of options o¤ered to would-be bond exchangers, and suggested that such opacity cast doubt on the eventual success of the swap.36 The sceptical reaction to the IIF “Financing O¤er” suggests to us that Euro 54bn was an unrealistic …gure, and should have seemed so at the time. Next, and equally if not more importantly, we turn to the issue of the net e¤ect of the PSI component on o¢ cial creditors’required commitments to Greece over the period 2011-2014. It is absolutely crucial to understand that, irrespective of the reduction in the value of the bonds to be amortized in the 2011-2014 period, PSI created novel …nancing needs that would not have otherwise occurred.37 One obvious cost would have been increased costs of recapitalizing Greek banks following the NPV losses on their huge holdings (relative to their size) of Greek government bonds. Such changes would
3 5 As mentioned in the text, confusingly the EC’ post-meeting Communique cites a much lower …gure of EUR 37bn. We s return to this below. 3 6 To give just one instance, on July 22 Goldman Sachs’reading of the IIF proposal was that “the expected notional amount of bonds involved will be in the region of EUR 13.5bn.” 3 7 All the points we are about to make were well-understood and widely discussed at the time. See, e.g., the July 2011 European Commission’ “Fourth Review” of the Greek programme (Occasional Paper 82), particularly Box 1. s

31

substantially increase the bank recapitalization bill and, since Greece was obviously unable to cope by itself, they correspondingly increased the …nancing bill for creditors in the o¢ cial sector. Possibly even more importantly, in order to induce private creditors to participate in the debt swap, the operation included a “credit enhancement”programme in which the new bonds issued by Greece would be guaranteed by AAA-rated bonds issued by the EFSF, i.e. the o¢ cial sector. This of course represents new commitments from o¢ cial creditors. There is confusion about the size of this new commitment. As mentioned, the EC’ Communique cites a contribution from the debt swap of EUR 37bn, potentially s suggesting an estimated cost of credit enhancement equal to EUR 54bn –37bn = 17bn. Subsequent to the meeting, however, the …gure most often cited for the cost of credit enhancement was EUR 35bn. These …gures suggest that credit enhancement alone reduced the net contribution of PSI by somewhere between 1/3 and 2/3 of the upper-bound …gure mentioned in the IIF document. Table A7 Here To get an example of some of the calculations circulating in the days after July 21st, we reproduce in Table A7 a Table from Eurobank Research (Greece Macro Monitor: August 2, 2011). It shows the EUR 54bn of …nancing coming from PSI (item B4 - note that they emphasize the assumption of 90% participation), but it also shows the increased …nancing need due to credit enhancement (a2) and Greek bank recapitalization (a4). Interestingly, a2+a4>B4, though we admit that not all the EUR 20bn in recapitalization needs can necessarily be attributed to PSI: the situation of Greek banks had deteriorated for other reasons as well.38 But bank recapitalization and credit enhancement are only two of several ways in which the debt swap –and PSI more generally –would have been likely to increase Greece’ overall …nancing needs –and hence s o¢ cial creditor commitments. Another way in which PSI almost certainly increased overall …nancing needs is by exacerbating Greece’ s inability to borrow from private creditors. Clearly Greece was struggling to persuade markets to buy its bonds for a number of reasons but, as we already argued in the text, the spectre of PSI had clearly been one of these (as testi…ed by many market participants’comments). From this perspective, the symbolic haircut contained in the July agreement is absolutely perverse. While it o¤ers virtually no relief to Greece
3 8 The table also shows another cost of the broad PSI component of the July agreement. i.e. EUR 20bn to …nance a debt-buyback operation. The debt buy back idea was another disaster, and it disappeared from the news fairly quickly. The reason is not hard to discern: …nancing the buyback would have required more money from the o¢ cial creditors, as is seen in the table. However the buyback is clearly distinct from the rollover, and we do not discuss it further.

32

and its o¢ cial creditors, it indicates to markets that further future haircuts are still on the table. This perpetuates Greece’ exclusion from private markets and correspondingly increases its need to rely on s o¢ cial creditors. Next, there are hard-to-quantify but plausibly signi…cant adverse general-equilibrium e¤ects of PSI on overall …nancing needs. In particular, to the extent that PSI worsened overall credit conditions in Greece (both by weakening Greek banks and by impairing market access for Greek companies), it also exacerbated the Greek depression, with obvious implications for the government de…cit and, once again, the need for support from EU partners and IMF. Finally, it is worth reiterating that PSI would have forced the acknowledgement of mark-to-market losses not only on Greek banks but also on banks in other European countries. In several of these countries the banking system was already in considerable stress, so it is not unlikely that some of the money allegedly saved by o¢ cial creditors through the bond swap would have had to be used to support their own banks. We acknowledge that it is very di¢ cult to quantify the overall impact on Greece’ …nancing needs from s these adverse consequences of PSI, but we believe it to be quite sizable. Coupled with our previous point that the euro 54bn of existing bonds whose maturity would be extended is almost certainly overestimated, we think that the net reduction in the burden for o¢ cial creditors is modest indeed, if not negative. Regardless, it seems inconceivable that such bene…ts would have outweighed the enormous potential costs (to the o¢ cial creditors themselves) of dragging Italy into the crisis.

33

References Alesina, Alberto and Silvia Ardagna, (2009): “Large Changes in Fiscal Policy: Taxes vs Spending,” Tax Policy and the Economy, Volume 24, 35-68. Alesina, Alberto and Alan Drazen (1991): “Why are Stabilizations Delayed?,” American Economic Review, Vloume 81, pages 1170-1188. Austen-Smith, David and Timothy Feddersen (2005): “Deliberation and Voting Rules,” In Social Choice and Strategic Decisions: Essays in Honor of Je¤ ery S. Banks, edited by David Austen-Smith and John Duggan, Springer-Verlag. Austen-Smith, D. and Timothy Feddersen (2006): “Deliberation, Prefrence Uncertainty, and Voting Rules,“ American Political Science Review, Volume 100(2), pages 209-217. Bini Smaghi, Lorenzo, “Challenges for the Euro Area, and the World Economy” Speech at The Group , of Thirty, 63rd Plenary Session, Rabat, 28 May 2010, pp. 3-5. Bini Smaghi, Lorenzo, “Eurozone, European crisis and policy responses” Speech at the Goldman Sachs , Global Macro Conference, Hong Kong, 22 Feb 2011, pp. 5-6. Borensztein, Eduardo and Panizza, Ugo, (2008): “The Costs of Sovereign Default,” IMF Working Paper 08/238. Buiter, Willem (2010a): “Sovereign debt problems in advanced industrial countries” Citi Economics, , Global Economics View, 26 April 2010, pp. 34-47. Buiter, Willem (2010b): "Greek sovereign debt restructuring delayed but not avoided for long” Citi , Economics, Global Economics Flash, 5 May 2010. Cline, William R. (2011): “Sustainability of the Greek Public Debt” Policy Brief 11-15, Peterson , Institute for International Economics. Cottarelli, Carlo, Lorenzo Forni, Jan Gottschalk, and Paolo Mauro (2010), “Default in Today’ Ads vanced Economies: Unnecessary, Undesirable, and Unlikely” IMF Sta¤ Position Note, September 1, 2010, , SPN/10/12. Coughlan, Peter J. (2000), “In Defense of Unanimous Jury Verdicts: Mistrials, Communication, and Strategic Voting,” American Political Science Review, Volume 94(2), pages 375-393. Cukierman, Alan and Tommasi, Mariano, (1998a): “When Does It Take Nixon to Go to China,”American Economic Review, Volume 88, pages 180-197. Cukierman, Alan and Tommasi, Mariano, (1998b): “Credibility of Policymakers and of Economic Reforms,” Federico Struzenegger and Mariano Tommasi, eds., The Political Economy of Reform, CamIn 34

bridge, MA: MIT Press. Drazen, Alan (2000): Political Economy in Macroeconomics, Princeton University Press. Drazen, Alan and Grilli, Vittorio, (1993): “The Bene…t of Crisis for Economic Reforms,” American Economic Review, Volume 83, pages 598-607. Eliaz, K…r, Ray, Debraj and Razin, Ronny, (2007): “Group decision-making in the shadow of disagreement,” Journal of Economic Theory, Elsevier, vol. 132(1), pages 236-273, January. Gerardi, Dino and Yarvin, Leeat, (2007): “Deliberative Voting,”Journal of Economic Theory, Volume 134, pages 317-338. Gerardi, Dino and Yarvin, Leeat, (2008): “Information Acquisition in Committees,” Games and Economic Behavior, Volume 62, pages 436-459. Hsieh, Chang-Tai, (1997): “Bargaining over Reform,” Working Paper. Lane, Philip R. (2012): “The European Sovereign Debt Crisis,”Journal of Economic Perspectives, 26, 3, 49-67. Lizzeri, Alessandro and Yariv, Leeat, (2011): “Sequential Deliberation,” Unpublished Meriowitz, Adam, (2006): “Designing Institutions to Aggregate Private Beliefs and Values,”Quarterly Journal of Political Science, Volume 1(4), pages 373-392. Orphanides, Athanasios (2010). Interview with Reuters, 19 April 2011, http://www.bis.org/review/r110805a.pdf . Orphanides, Athanasios (2011): "New Paradigms in Central Banking?” Working Papers 2011-6, Cen, tral Bank of Cyprus. Perotti, Roberto (2011): “The Austerity Myth,” unpublished, Bocconi University. Persico, Nicola, (2004): “Committee Design with Endogenous Information,”Review of Economic Studies, Volume 71(1), pages 597-612. Ponsati, Clara and Sákovics, Jozsef, (1995): “The War of Attrition with Incomplete Information,”Math. Soc. Sci., Volume 29, pages 239-254. Reinhart, Carmen, and Kenneth Rogo¤ (2011): This Time Is Di¤ erent: Eight Centuries of Financial Folly. Roubini, Nouriel (2010): “Greece’ Best Option is an Orderly Default” Financial Times, 28 June 2010, s , http://www.ft.com/cms/s/0/a3874e80-82e8-11df-8b15-00144feabdc0.html . Zettelmeyer, Jeromin, Christoph Trebesch, and Mitu Gulati (2013): “The Greek Debt Exchange: An Autopsy,” Economic Policy. 35

Table 1: Greek package signed in May 2010 Primary deficit/GDP 2009 2010 2011 2012 2013 2014 2015 -8.6 -2.4 -0.9 1 3.1 5.9 6 Total deficit/GDP -13.6 -8.1 -7.6 -6.5 -4.9 -2.6 -2 Gov. debt/GDP 115.1 133.2 145.2 148.7 149.2 146.1 140.4 Real GDP growth (%) -2.0 -4.0 -2.6 1.1 2.1 2.1 2.7 250 200 150 100 100 100 Spread over bunds T-bill issuance (Eur bn) 6.1 8.0 8.0 8.0 8.0 8.0 8.0 Bonds issuance (Eur bn) 55.6 25.4 4.4 23.4 34.9 64.5 66.8

Source: IMF May 2010, Country Report No. 10/110

Figure 1: Negotiation with Non‐Single Peaked Preferences
ECB Loss Function

German Loss Function

Bargaining Outcome

Germany’s Bliss Point

Size of Haircut

Table A1: % point reduction of primary deficit/GDP in largest episodes of fiscal adjustments in OECD countries and in Greek programs May-10 NLD 96-00 T T+1 T+2 T+3 T+4 T+5 T+6 Mean Cumulative 1.9 9.7 7.3 0.4 0.2 0.9 0.9 BEL 84-90 4.6 0.9 0.4 1.7 0.5 1.0 1.2 1.5 10.2 UK 94-00 1.3 1.4 1.7 2.0 2.0 0.5 2.6 1.7 11.6 4.2 17.0 2.5 12.4 1.7 8.7 2.2 6.7 2.4 11.8 1.9 9.3 2.4 14.6 DNK 83-86 4.0 5.6 2.5 4.8 SWE 94-98 2.9 1.1 4.4 2.0 2.1 CAN 93-97 0.4 1.9 1.9 2.1 2.4 IRE 87-89 2.0 3.3 1.3 FIN 94-98 GRC 90-94 GRC 10-15 3.0 0.4 3.2 2.5 2.7 0.4 3.8 0.5 -0.1 4.7 6.2 1.5 1.9 2.1 2.8 0.1 Jul-11 GRC 10-15 5.4 4.0 2.4 2.3 2.1 2.8 1.3 3.0 18.0

Source: OECD Economic Outlook and IMF

Table A2: Realized and forecast real GDP growth rates in the largest episodes of fiscal adjustments versus Greek programs May‐10 Jul‐11 GRC 10‐15 GRC 10‐15 ‐4.0 ‐4.5 ‐2.6 ‐3.9 1.1 0.6 2.1 2.1 2.1 2.3 2.7 2.7

T T+1 T+2 T+3 T+4 T+5 T+6 Total Growth (Index, T‐ 1=100) Average Ann. Growth

NLD 96‐00 BEL 84‐90 3.4 2.1 4.3 1.8 3.9 1.9 4.7 2.4 3.9 4.6 3.6 3.1 121.9 121.1

UK 94‐00 4.3 3.1 2.9 3.4 3.8 3.7 3.9 127.9

DNK 83‐86 SWE 94‐98 CAN 93‐97 IRE 87‐89 2.7 4.0 2.3 3.6 4.2 3.9 4.8 3.0 4.0 1.6 2.8 5.6 4.9 2.7 1.6 4.2 4.2

FIN 94‐98 3.6 4.0 3.6 6.2 5.0

GRC 90‐94 0 3.1 0.7 ‐1.6 2

116.7

117.6

116.8

112.7

124.5

104.2

101.2

99.0

4.0

2.8

3.6

3.9

3.3

3.2

4.1

4.5

0.8

0.2

‐0.1

Shade cells denote IMF forecasts Source: IMF

Table A3: Output Gap (% of potential output) change in the largest episodes of fiscal adjustments versus Greek programs May‐10 Jul‐11 GRC 10‐15 GRC 10‐15 4.0 4.2 ‐3.1 ‐0.4 ‐1.4 ‐7.1 ‐0.9 ‐4.6

Start Finish Average Ann. Change Total Chang

NLD 96‐00 BEL 84‐90 ‐3.3 ‐1.1 2.5 2.2 1.2 5.8 0.5 3.3

UK 94‐00 ‐2.4 0.7 0.4 3.1

DNK 83‐86 SWE 94‐98 CAN 93‐97 IRE 87‐89 7.2 ‐5.4 ‐1.9 ‐1.6 15.0 ‐0.9 ‐1.2 ‐0.7 2.0 7.8 0.9 4.4 0.1 0.7 0.3 0.9

FIN 94‐98 ‐6.8 1.6 1.7 8.5

GRC 90‐94 2.2 ‐1.1 ‐0.7 ‐3.4

Source: IMF

Table A4: Maximum general government gross debt (% GDP) in the largest episodes of fiscal adjustments versus Greek programs May‐10 Jul‐11 GRC 10‐16 GRC 10‐16 149.2 172 2013 2012

Maximum Year

NLD 96‐00 BEL 84‐90 74.1 125.6 1996 1990

UK 94‐00 49.3 1997

DNK 83‐86 SWE 94‐98 CAN 93‐97 IRE 87‐89 71.9 73.2 101.7 109.2 1984 1996 1996 1987

FIN 94‐98 56.5 1994

GRC 90‐94 100.5 1993

Source: IMF

Table A5: % point change of the current account in the largest episodes of fiscal adjustments versus Greek programs NLD 96‐00 BEL 84‐90 Current Account Balance at the start of the adjustment (%GDP) T T+1 T+2 T+3 T+4 T+5 T+6 Cumulative 6.2 ‐1 1.4 ‐3.3 0.6 ‐1.9 ‐0.6 0.5 0.9 1.8 ‐0.7 0.3 0 ‐0.4 2.4 UK 94‐00 ‐1.9 0.9 ‐0.3 0.4 0.7 ‐0.3 ‐2 ‐0.3 ‐0.7 DNK 83‐86 SWE 94‐98 CAN 93‐97 IRE 87‐89 ‐3.9 1.7 ‐0.8 ‐2.4 ‐0.6 ‐1.3 2.4 2.2 0.2 0.6 ‐0.3 ‐3.6 ‐0.2 1.6 1.5 1.3 ‐1.8 ‐3 2.9 0.2 ‐1.5 FIN 94‐98 ‐1.3 2.4 3 ‐0.1 1.6 0 GRC 90‐94 ‐3.4 ‐0.4 2.3 ‐0.4 1.2 0.6 May‐10 Jul‐11 GRC 10‐15 GRC 10‐15 ‐11.2 2.8 1.3 1.5 1.6 1.2 0.9 9.3 ‐11 0.6 2.2 1.1 0.8 0.3 0.8 5.8

‐4.3

‐2.1

5.1

2.3

1.6

6.9

3.3

Source: IMF

Table A6: Summary of Real Effective Exchange Rates (REER) based on CPI and monetary regimes in the largest episodes of fiscal adjustments versus Greek program Country NLD 96‐00 BEL 84‐90 UK 94‐00 DNK 83‐86 SWE 94‐98 CAN 93‐97 IRE 87‐89 Change in REER During Adjustment [1] Depreciated by 12. 9% Appreciated by 7.1% Appreciated by 27.2% Appreciated by 9.6% Depreciated by 10.9% Depreciated by 21.5% Depreciated by 3.6% Exchange Rate Regime Pegged in ERM, fixed from 1998 Pegged in ERM Floating, abandoned peg Pegged in ERM Floating, abandoned peg Floating Pegged in ERM Notes The REER of the Dutch Guilder remained broadly stable during the 2 years immediately before the fiscal adjustment. The Belgian Franc devalued by 8.5% in February 1982 against EMS currencies. The UK abandoned the peg in September 1992 resulting in the REER devaluing by 15% from 1992Q3 to 1993Q1. The Danish Krone did not devalue its peg but its REER did depreciate by 7.3% in the 2 years before the adjustment. Sweden abandoned the peg in November 1992 resulting in the REER devaluing by 22.0% from 1992Q3 to 1993Q3. The REER of the Canadian Dollar fell modestly by 2.4% in the 2 years before the adjustment started. The Irish Pound devalued by 8% in August 1986 against EMS currencies but this had a muted effect on Ireland’s REER as the British Pound was depreciating significantly at the time. The Finnish Markka devalued by 12% against EMS currencies in November 1991 before abandoning the peg in September 1992 resulting in the REER devaluing by 16.7% from 1992Q3 to 1993Q3. This era of Greece’s economy is marked with high but declining inflation which peaked in 1990 at 27% before decreasing to 11% by 1994. The REER of the Drachma appreciated modestly by 4.3% during the 2 years immediately before the adjustment. The REER of Greece had appreciated by 19 percentage points since joining the Euro in 2002. Figure 2 shows the divergence in competitiveness based on CPI between Greece and it's main trading partner Germany in the decade before the first bailout. Cumulative Current Account Change (% GDP) ‐4.3 2.4 ‐0.7 ‐2.1 5.1 2.3 1.6

FIN 94‐98

Appreciated by 10.6%

Floating, abandoned peg

6.9

GRC 90‐94

Appreciated by 14.2%

Floating ?

3.3

May‐10 GRC CPI Inflation was forecast Eurozone 10‐15 to remain relatively stable vis a vis German inflation (see figure 4)

9.3

[1] REER data is derived from the nominal effective exchange rate using the quarterly release of CPI data. The series are produced by the IMF and obtained via Thomson Reuters DataStream. The series have a base year 2005=100 and change "during adjustment" referes to the value of the series in the first quarter of the first year of the adjustment to the value of the series in the last quarter in the last year of the adjustment. Historical data of REERs based on ULCs have also been obtained and display similar changes. No forecast of Greece's REER over the adjustment period is available from the IMF. Verification that Greece had a floating exchange rate regime (90‐94) is needed.

Table A7: Reproduced from Eurobank Research (Greece Macro Monitor: August 2, 2011)

150 140 130 120 110 100 90 80

Figure A1: Realized and Forecast CPI Inflation Index 2000 = 100 (2000 - 2015)

Greece Germany

For 2010-2015 the figure shows IMF forecasts as of May 2010. Sources: All German data (with the forecasts) and Greek data up to and including 2009 is from the April 2010 IMF WEO Report. Forecasts for Greek inflation 2010‐2015 are from the May 2010 IMF Country Report No. 10/110.

2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015