Negotiating implementation of EU fiscal governance

ABSTRACT Since 1997 the Economic and Monetary Union has been centred on the Stability and Growth Pact (SGP), which provides the framework for the coordination of economic policies and the surveillance of fiscal policies of member states. The country-specific recommendations, which address the member states’ macroeconomic programmes, are the policy instruments of the SGP preventive arm. At the same time, these recommendations are the object of intense negotiation between the Commission and the Council. Why are these recommendations a matter of bargaining between the Commission and the Council? What are the factors that determine the negotiation outcomes? Employing a provision-by-provision text analysis, I show that the Council strengthens almost 50 percent of the country-specific recommendations proposed by the Commission between 1999 and 2015. The Council tightens the recommendations during the financial crisis and against highly indebted countries. Recommendations addressed to large countries are instead less likely to be tightened.

and during the sovereign debt crisis. In line with Baerg and Hallerberg (2016) recommendations addressed to large countries, which exert greater bargaining power in the Council, are more likely to be watered down.

Fiscal rules and the country-specific recommendations
The surveillance mechanism of macroeconomic programmes coordinates and monitors fiscal policies of EU member states. According to article 140 of the Treaty on the Functioning of the European Union and the relative Regulations of the SGP and Six Pack, each member state is required to submit annually to the Commission for evaluation a stability or convergence program, 1 which includes detailed information on the economic and budgetary situation. The successive regulations on the preventive arm have become more specific about the information to provide in these programmes. The 1997 regulation referred to medium-term objectives, information on deficits and debts, general description of the budgetary and other economic policy measures and the main assumptions about expected economic developments (articles 3 and 7). In the 2005 reform, the programs replaced the common medium term objective (MTO) with a country-specific objective which could diverge from a close-to-balance or surplus position, and comprised detailed and quantitative assessments of the budgetary and other economic policy measures being taken or proposed to achieve, including an accurate cost-benefit analysis of major structural reforms. Finally, the Six Pack (2011) adds the expenditure benchmark as well as information on implicit liabilities related to ageing and on the consistency of the macroeconomic programmes with the broad economic policy guidelines and the national reform programmes.
After examination, the Commission writes an in-depth report on each national program and proposes recommendations to the Council, which adopts them by qualified majority voting. Negotiations between the Commission and the Council and among member states are example of majority-rule bargaining under open rule (Baron and Ferejohn 1989). The Commission uses agenda-setting power to write the report, but the adoption of this document may be hampered by the Council's open rule. On the other hand, the proposal power of the Council is mitigated by the supermajority voting. Each minister may propose amendments, which require a qualified majority for approval.
At the core of these policy instruments are the key priority issues that require member states' immediate attention. They deal with not only the fiscal policies, as established by the SGP, but also those policies inserted by the Six Pack and aimed at redressing macroeconomic imbalances (in accordance with Regulation 1176/2011 and Regulation 1174/2011) and addressing other major economic objectives, such as growth enhancing structural reforms, employment and social aspects and/or financial market stability. Member states are expected to follow these recommendations when they draft their national budget for the following year. This surveillance mechanism has involved every member state, with the exception of those subject to an Economic Adjustment Programme, 2 and it has been part of the new coordination system of the European Semester since 2011. 3 While Baerg and Hallerberg (2016) have investigated only the overall assessments produced under the SGP, they have ignored the politically binding nature of the country-specific recommendations. A failure to implement these recommendations might ultimately result in sanctions under the SGP and the macroeconomic imbalances procedure. In other words, I argue that the Commission and the Council bargain over the detailed measures to be implemented at national level. I illustrate this point considering both the propensity and the nature of change of these recommending provisions for the entire period SGP has been in operation.
Bargaining over compliance with fiscal governance rules

Commission-Council conflicts
The relationship between inter-and intra-institutional bargaining and policy change is at the core of several formal models of both EU and the US decision-making. A large part of this literature is primarily focused on explaining policy gridlock and policy change across several legislative systems. The basic concepts of the pivot theory, namely the gridlock interval and the pivot actor, have been applied also to the EU political system (Citi and Justesen 2016;Crombez and Hix 2015;Junge, König, and Luig 2015). The gridlock interval indicates the set of policies that cannot be changed under a given decisionmaking rule. While under simple majority, the median voter is usually the pivotal actor, that may not be the case with qualified majority voting.
Consider a left-right unidimensional policy space and member states with preferences p distributed along this space. The state with preference p L , which is pivotal for a rightward policy move, can form a blocking minority with states on its left of the policy space, while the state with preference p R , which is pivotal for a leftward policy move, can form a blocking minority with states on its right of the spectrum. The resulting gridlock interval, GI = [p L , p R ], is the set of policies that cannot be modified. In view of its agendasetting power, the Commission may advance policy recommendations that departs from the status quo. Specifically, it may make proposals that are attractive to the pivotal members and preferred to the status quo (Citi and Justesen 2016). These policy recommendations are less likely to be amended by the Council when the Commission proposes a policy located inside the gridlock interval. In other words, the wider the gridlock interval, the lower the likelihood of policy change.
The gridlock interval however cannot help us explain the direction of change, that is the tightening or weakening of policy recommendation. Liberal intergovernmentalism (Moravcsik 1998) and spatial models of EU decision-making (Crombez 1997(Crombez , 2001Garrett 1995;Garrett and Tsebelis 1996;Steunenberg 1997;Tsebelis and Garrett 2000), instead, suggest that the location of the status quo and the distribution of actors' preferences should affect the bargaining outcome. In negotiations over the countryspecific recommendations, the status quo or best alternative to a negotiated agreement is no recommendation, which would clearly leave complete autonomy to member states on how to conduct their macroeconomic policies. Governments that would benefit the most from this situation, and therefore enjoy greater bargaining power, are those facing greater constraints from their public finances since they are likely to face demanding policy recommendations.
Historically in Europe, the ideological and partisan orientation of governments influences national preferences on macroeconomic policies (Alesina, Roubini, and Cohen 1997). Leftist parties seek to keep unemployment permanently low and adopt expansionary social policies. On the contrary, right-wing parties prefer to provide fewer social programs and more tax cuts. This explicit gap may ultimately lead to different fiscal preferences once they are in power. Alesina, Roubini, and Cohen (1997) allege that the propensity of left-wing governments to support social policies increases the risk of fiscal profligacy and higher debt. Because of their macroeconomic policies, I assume that in negotiations over the country-specific recommendations the Commission would impose greater policy demands to left-wing countries, in order to enforce fiscal discipline and prevent noncompliance.
But what can left-wing governments do to soften the Commission's recommendations? One factor could work in their favour: the position of the pivotal government. Once the Council receives the Commission's proposal, the pivotal member will only accept policies it prefers to the status quo. The threat of non-agreement (Moravcsik 1998, 63-64) guarantees that the state with preference p L would gather a winning coalition and soften the Commission's recommendations. I expect that the closer the left pivotal country is to the status quo and the more likely the Commission's recommendation is weakened.

Country power
Ministers within the Council have the incentive to negotiate their most preferred policy and the outcome of these negotiations mostly depends on the asymmetries of countries' bargaining power. Bargaining outcomes tend to be closer to positions of the most resourceful states (Thomson et al. 2006). Voting weights and the size of the economy are common proxies for bargaining resources (Bailer 2004;Thomson 2011). Governments from larger and smaller countries may hold different bargaining power in these negotiations. Early on in the procedure, bargaining over fiscal policy recommendations involves primarily the Commission. The supranational institutions is composed by a College of Commissioners, one for each member state, 4 voting by simple majority. This means that smaller countries are overrepresented in relation to their population size. Several works have found that smaller states enjoy more bargaining success because of the Commission's influence on decision outcomes (Thomson 2008), narrower interests and clearer priorities (Arregui and Thomson 2009;Bailer 2004). Countries with higher voting weight have higher probability of extracting better concession within the Council where they have more possibilities to set alternative majority coalitions and have more potential for compromise and linkage (Moravcsik 1998;Thomson et al. 2006). Applying this argument, powerful states are more likely to water down the Commission's recommendations because of the influence they exert in the Council (Baerg and Hallerberg 2016). Therefore, I expect that larger countries having greater voting weights are more successful in weakening the Commission's recommendations.
The influence of economic size is less clear-cut. Bailer's (2004), Bailer, Mattila, and Schneider (2015) and Tallberg's (2008) recognize that economic size has a positive and significant effect on Council negotiation. Larger countries are more likely to secure a more favourable treatment because of their greater negotiating abilities and issuing more credible threats towards insolvent countries (Heipertz and Verdun 2010;Schimmelfennig 2014;Talani and Casey 2008). Heipertz and Verdun (2010) find that countries with large economies were more successful in reforming the SGP in 2005. Baerg and Hallerberg (2016) affirmed that economic performance matters for Council's decision to amend the Commission's recommendations. Unfortunately, corroborating evidence is hard to come by.
This logic of intergovernmental bargaining based on macroeconomic situation has been extended to EU financial transfers. The literature on coalition-building in the Council finds evidence of a cleavage between net-contributors and net-beneficiaries in the EU budget through structural and agricultural funds (Bailer, Mattila, and Schneider 2015;Carrubba 1997;Zimmer, Schneider, and Dobbins 2005). In the wake of member states' financial difficulties that arose from the debt crisis, a wide array of financial support mechanisms were established. 5 Net-beneficiaries as well as those countries subject to financial assistance (i.e. Greece, Hungary, Ireland, Latvia, Portugal, Romania and Spain) may accept less demanding recommendations for contributing countries in exchange of financial support, since penalizing countries that contribute the most to the EU budget would have adverse effects on the poorer countries and on the EU-wide economic situation (Hansen 2015). I leave to the empirical section the corroboration of these propositions.

Compliance records with the Stability and Growth Pact
The EU literature on fiscal compliance is mainly focused on the corrective arm, identifying a number of reasons why compliance with the two reference values has faltered 6 (Beetsma and Giuliodori 2010; Beetsma and Uhlig 1999;De Haan, Berger, and Jansen 2004;Hansen 2015;Heipertz and Verdun 2010;Koehler and König 2015;Varela and Prado-Dominguez 2012). However, these researches disregard the constraints imposed by the preventive arm, with the sole exception of Baerg and Hallerberg (2016). Most importantly for my purpose, the literature ignores how countries' debt and deficit level affect bargaining over the country-specific recommendations. 7 In the negotiation over fiscal implementation, the Commission, as guardian of the treaties, would reward the fiscally sound countries and demand tougher policy actions from countries with high debt and excessive deficit. As a response, governments may decide either to weaken or to strengthen the recommendations, depending on the economic costs and the reputational costs they would incur. Consider a noncompliant country and their fellows in the Council. Governments would engage in logroll if they obtain satisfactory concessions from their colleagues in return, either on other issues of the same legislation or on other legislations (Aksoy 2012). In negotiations over different country-specific recommendations, fellow governments at (future) risk of noncompliance have higher incentive to weaken the Commission's recommendations, because they would mutually gain as much in the case they run in the same situation. I expect that when a country is noncompliant, the fellow governments are more likely to weaken the Commission's recommendations, because they would benefit from logrolling.
On the other hand, a reputation-based argument could lead to the opposite explanation. The SGP was created in order to increase budgetary discipline, enhance credibility of fiscal governance and reduce governments' discretion on their short-term actions when facing economic shocks (Beetsma and Giuliodori 2010;Heipertz and Verdun 2010). Any attempts of undermining the Commission's assessments and, ultimately, breaking the SGP would be a stigma for the proposing countries. In the early years of the SGP, indeed, the loss of political reputation was significant for countries breaching the deficit limit, but not for countries breaching the debt limit (von Hagen and Wolff 2006;Hansen 2015). In negotiations over country-specific recommendations, fellow governments would refrain from supporting noncompliant behaviour, otherwise they would incur in reputational losses as well. I expect that when a country is noncompliant, the fellow governments are more likely to leave unchanged or strengthen the Commission's recommendations, because they would lose their reputation.
To sum up, two alternative causal mechanisms explain the influence of the record of compliance with the fiscal rules over adopted recommendations: a logrolling and a reputation-based argument. I consider both for empirical evaluation.

Economic distress
Although the deficit and debt reference values have been crucial criteria used by the Commission to judge whether countries met the requirements to join the Euro Area and to be compliant with the SGP, they have hidden the existence of significant differences among countries in respect to their economic situation and their position in the business cycle. Germany and norther countries have pursued a supply-side or export-led growth strategy based on wage restraint, high value-added goods, and competitiveness of national firms on international market. By contrast, southern countries have pursued a demand-led growth strategy based on fiscal expansion and wage inflation (Hall 2012). In a monetary union with a central bank, these different strategies resulted in higher interest rates, currency appreciation, trade imbalances, and high vulnerability to the financial market turbulence for southern countries. The divergence of these strategies intensified in 2010, when the outbreak of the sovereign debt crisis, triggered by the worldwide financial crisis, put under severe stress the European public finances, Greece required the first bailout and also other countries experienced contagion, in particular Ireland, Portugal and Spain. In period of economic distress, the Commission as well as the Council should be willing to do whatever it takes to limit this risk of sovereign default, breakdown of the monetary and financial system and a long-lasting recession (Schimmelfennig 2015). In being determined to avoid contagion, the Council would be more likely to accept Commission's proposals to initiate an EDP, under the corrective arm, and to strengthen the surveillance and coordination instruments of the preventive arm. Therefore, I expect that, when a country is facing economic downturn, the Council would both amend and strengthen the Commission's recommendations.

Domestic political constraints
The last factor is the domestic constraint derived from public attitudes towards the EU. Supported by a pro-European domestic public the Commission can bolster its credibility as monitoring and sanctioning actor. Ministers, on the other hand, are re-electionminded politicians who are concerned about the outcomes of their supranational decisions. In response to stringent or lax recommendations, their constituencies may ultimately decide whether to reward or punish their governments at the next electoral rounds. This factor has influential roots in Putnam's (1988) article on two-level games and its role has been largely recognized in the literature on EU bargaining over both legislation and treaties (Bailer and Schneider 2006;Hagemann, Hobolt, and Wratil 2017;Hosli 2000;Hug and König 2002). Applying Putnam's domestic win-set to EMU bargaining, Hosli (2000) shows that the risk of rejection by the domestic politics in Germany put in serious risk the acceptance of the intergovernmental agreements about the European economic governance. National representatives within the Council may signal their positions towards the EU integration and may change them in line with the public opinion. Governments are more likely to oppose legislative proposals when their domestic electorates are sceptical about the EU (Hagemann, Hobolt, and Wratil 2017). In their work on the negotiation over country-specific recommendations, Baerg and Hallerberg (2016) find empirical evidence that countries with low public support for the EU are more successful in weakening Commission's assessments. Country size reinforces the positive relationship between Euroscepticism and the probability of weakening the Commission's recommendations.
I expect that when the public opinion has pro-European attitudes, governments react accordingly, that is they are less likely to amend and more likely to strengthen the Commission's proposals.

Data collection and text analysis
In the literature, there are essentially two methods to measure interinstitutional conflicts: expert survey and text analysis. The first one has been at the basis of influential contributions over the EU legislative politics (e.g. Thomson et al. 2006;Thomson 2011), none with the country-specific recommendations in mind, though. Expert survey has the limit of how far back in time you go because experts' memory fades and their availability diminishes. Baerg and Hallerberg (2016), instead, employ text analysis. They use the Commission's recommendation and the following Council's opinion to produce a dichotomous measure of the direction of changes adopted by this latter institution in the country-specific recommendations. 8 In this study, I follow the text analysis approach and develop a new coding scheme, which tracks the substantive interinstitutional disagreements when giving recommendations to the member state. I rely on 395 country-specific recommendations 9 released between 1999 and 2015. Because I cannot produce a measure of compliance bargaining when the country is subject to Economic Adjustment Programme, the final dataset comprises 376 country-specific recommendations.
To measure interinstitutional conflicts over the country-specific recommendation I compare the Commission's recommendation with the Council's opinion on the stability and convergence programmes. 10 I broadly follow the method developed by Baerg and Hallerberg (2016), using hand-coding text analysis, but with some expedients being able to capture the substantive interinstitutional disagreements when giving recommendations to the member state. Although Baerg and Hallerberg's procedure allows analysing the full history of negotiations over the country-specific recommendations, it has some problematic features. First, their operationalization is not fully convincing because the level of analysis is the entire document rather than those provisions demanding policy actions to the countries. Indeed, these provisions are considered the linchpins of the preventive arm. Second, they do not explain how they have aggregated the estimates for each document.
In this study, the two dependent variables -Council change and Council severityare measured at the level of a given country-specific recommendation (csr). Council change counts the number of recommending provisions that have been modified by the Council. Each modification is then coded as either 'strengthened' or 'weakened'. Council severity is an ordinal variable, which measures the direction of change as follow: If more provisions have been weakened than strengthened, it takes the value of −1, otherwise, it takes the value of 1. If provisions have not been changed or the countryspecific recommendation contains the same amount of strengthened and weakened provisions, it takes the value of 0. 11 In Appendix 1 I explain at length the coding rules used to produce these data. In Appendix 2 I discuss the validity of these estimates comparing them with those produced by Baerg and Hallerberg and I run robustness checks with alternative dependent variables. I consider the following explanatory variables. I adopt the variables developed by Crombez and Hix (2015) to measure the Gridlock interval, 12 which is the absolute difference in preference between the left and right pivotal governments. A wider gridlock interval indicates narrower opportunities for policy change. If the Commission issues attractive proposals to member states, the recommendations are less likely to be amended. The Commission position takes the value of one if the median Commissioner is located outside the gridlock interval. 13 In the analysis of the direction of change, the closer the left pivotal government is and the higher is the likelihood that the Council softens the recommendations. To explain the tightening or weakening of policy recommendations by the Council, I include the variable Pivot that is the distance of the left pivotal government from the status quo.
As far as the member state power is concerned, amendments and weaker constraints imposed by the Council should be more likely when the country is more politically or economically powerful. The variable Political Power is measured by the proportion of the voting share each country enjoy in the Council qualified majority voting rule. This index considers the changes resulting from two voting systems (before and after the Treaty of Nice) and from subsequent EU enlargements. For example, in 2000 Germany had 11.5 percent of votes, while ten years later it carries 8.4 percent of votes. In order to capture Economic Power, the logarithm of the GDP is given for each country for each year. 14 Since economic resources in the EU are allocated from solid countries to poorer countries, I consider this variable a reasonable proxy for the national contribution to the EU budget.
The expected influence of compliance records on the negotiations over countryspecific recommendations is measured by the public debt and public deficit as share of GDP. Both Debt/GDP and Deficit/GDP are reference values for SGP compliance.
Economic differences among countries in terms both of fiscal positions and of the impact of the sovereign debt crisis are captured by two business cycle variables: GDP growth and Inflation rate. Moreover, to account for the sovereign-debt crisis I include a dichotomous variable, Economic crisis, that takes the value of one if the country-specific recommendation is adopted in the period between the first Greek bailout in 2010 and 2015.
Finally, governments are more likely to weaken the Commission's recommendations when facing public opposition towards the EU. From the Eurobarometer survey, I employ the question asking respondents whether they think that their country's membership is a 'good thing', a 'bad thing' or neither. 15 I measure the variable Public opinion as the share of respondents answering that membership is good, net of the share of respondents answering that membership is bad. Table 1 provides a summary of these variables with some descriptive statistics.

Outcomes of negotiations over country-specific recommendations
The Council modifies more than two thirds of the country-specific recommendations. Table 2 (2016), the Council has strengthened almost half of country-specific recommendations (187 out of 376), rather than weakened (12.5 percent), or left unchanged (37.6 percent). With the exception of four years (in 1999, 2000, 2002 and 2006), the Council has been stricter than the Commission over the entire period under examination, as Figure 1 shows. The most demanding modifications of the Council occurred in 2010 when 81 percent of proposed Commission's recommendations were strengthened. The Council's strengthening effort is confirmed even by looking at the distribution by member states. There is a clear variation, though. The most strengthened countryspecific recommendations are addressed to France (11 out of 17 documents), Greece (10 out of 12, even though the country has been subject to Economic Adjustment  Programme since 2011), followed by Belgium and Germany (9 out of 17 recommendations), Hungary and Slovenia (9 out of the 12 recommendations after the 2004 enlargement). Among the ten founding member states of the Eurozone, only Luxembourg, Finland and Ireland have not been subject of more vigorous constraints by the Council for at least half of the annual recommendations. This is not surprising, though, given the most demanding fiscal rules applied to Eurozone countries. More interesting for my purpose, the smaller and more fiscally virtuous countries, such as Luxembourg, Finland and Denmark (although outside the Eurozone), have the highest number of countryspecific recommendations. The first part of the inferential analysis examines what motivates the Council to modify the Commission's proposal. Because Council change is an over-dispersed count variable, I use a negative binomial regression with country and year fixed effects. The second part deals with the weakening or strengthening of the country-specific recommendations and I run an ordinal logistic regression with country and year fixed effects (Long and Freese 2014). Table 3 reports the results of these two models, testing both for the entire period and for the period before the Six Pack.
Inter-institutional conflicts do not seem to operate as the pivot theory suggests. Both the size of the gridlock interval and the position of the Commission are statistically insignificant. The expectation that the closer is the left pivotal government to the status quo and the more likely the initial text is softened is not corroborated.
Country power partially matters. Countries with higher voting weight have higher probability not only of amending the Commission's proposal, but also of extracting better concessions from the fellow governments. Confirming Baerg and Hallerberg's (2016) results, there is strong evidence that recommendations addressed to large countries are more likely to be weakened, rather than strengthened, by the Council. In 2007 Germany and Luxembourg held 29 and 4 votes, respectively. If Germany, instead of Luxembourg, were under scrutiny, the predicted probability of weakening the Commission's recommendations is between 13.4 and 14.3 percent, instead of 9 and 7.5 percent (for the entire and the pre-Six Pack period, respectively). On the contrary, there is weak evidence that countries with strong economies are more likely to keep the Council in check and to receive more accommodating recommendations.
Expectations on compliance partially confirm the reputation-based explanation. The probability to amend the Commission's proposals by the Council increases by 118.4 percent, if the country under scrutiny runs a large fiscal deficit (i.e. such as Ireland in 2010 with a 32.3 percent deficit/GDP ratio) instead of being compliant (i.e. Germany in 2015 had a surplus of 0.7). As far as highly indebted countries are concerned, the dynamics is more interesting. They are likely to be subject to more strengthening recommendations by the Council, especially in the period before the entry into force of the Six Pack. The highest indebted countries were Greece in 2010, with more than 146 percent debt-GDP ratio, and Italy in 2015, with 132.3 percent ratio, while Estonia in 2007 had the smallest ratio. The Council is 80.2 percent more likely to modify a recommendation addressed to Greece than to Estonia. During the entire period, the chances that the Council strengthens the recommendations addressed to Italy, compared to Estonia, increase of 26.2 percent. In the pre-Six-Pack period, the chances that the Council amends and strengthens the Commission's recommendations are even higher. The results show that strengthening recommendations are 41.3 percent more likely in the case of Greece  than Estonia. A reasoning by backward induction may be at stake here. Given that before 2011 the debt criterion has been largely ignored in the implementation of the corrective arm, that is in the opening up of the EDP, the Commission and the Council have the incentive to better scrutinize and make stronger recommendations in the preventive phase when assessing highly-indebted countries in that period. Evidence from the economic distress expectations is weaker. There is strong evidence that in the years following the outbreak of the economic and financial crisis, the Council is more active, perhaps because of concerns about contagion it favors more austere policy measures. After 2010 the probability of amending increases by between 40 and 64.9 percentage points and the probability of strengthening increases by between 10.4 and 30.9 percentage points. Moreover, business cycle variables present mixed results. If a country has a high inflation rate of 4.2 percent (90 th percentile), instead of reaching the price stability criterion defined by the European Central Bank (inflation rate below but close to 2 percent), the probability of Council's amendment increases by between 19.5 and 16.9 percentage points. Evidence on the GDP growth is weak but confirms the expectation that the Council is more accommodating towards countries facing favorable economic situation.
Finally, domestic political constraints do partially matter and confirm the hypothesis. A move from the least Eurosceptic country, which is Ireland in 2000 (83.12 percent of respondents supporting the European integration), to the most Eurosceptic country, which is Czech Republic in 2011 (30.53 percent of respondents supporting the European integration), decreases the probability of amending the Commission's proposal by between 34.4 and 41.9 percentage points. In line with Baerg and Hallerberg (2016) findings, the Council is more tolerant when the subjected country suffers high level of Euroscepticism. This results are, however, not statistically significant.

Conclusion
In this article, I explore one policy instrument designed to prevent incompliance with the EU fiscal rules and to foster cooperation between the national and supranational actors. I draw expectations on the negotiation over the implementation of the EU policy recommendations under the preventive arm. One might think that the procedure gives a strategic advantage to the Council and this would lead to a softening of the countryspecific recommendations. In light of the literature on EU bargaining and compliance, this study challenges this intuition. I test for the factors affecting modifications of Commission's proposals with an innovative text analysis of Commission's recommendations and Council's opinion. Counterintuitively, evidence shows that the Council has managed to amend 71 percent of Commission's proposals. Moreover, the Council has more likely strengthened rather than weakened such recommending provisions, but not all member states have been penalized. Large countries, indeed, may take advantage of their voting power and, despite being targeted in the early stage by the Commission, they can easily form coalitions, at least in the whole period, to amend and weaken the Commission's recommendations.
From the standpoint of the Commission, this study offers also solace. The Council cannot overlook when the country under scrutiny exceeds either the deficit or debt reference value and it is inclined to be more active in recommending stricter policy actions. Surprisingly, measures addressed at highly indebted countries are more likely to be strengthened before, rather than after, the latest reforms. A reasonable explanation lies on the different prominence given to the debt criterion in the preventive and corrective arm, before the Six Pack. Since the debt criterion was largely ignored in the decision to open up an excessive deficit procedure, the Commission and the Council may have had the incentive to assess the highly indebted country's programs more strictly in that period.
Under financial distress, fear of disintegration and contagion drives the Council to avoid conflict with the Commission. National governments, ultimately politicians, cannot ignore the danger of the sovereign debt crisis and are obliged to follow, and even strengthen, the Commission's recommendations. These results, however, need further analysis, considering that economic growth after the crisis is still highly heterogeneous and this may have an impact on the inter-institutional conflicts as well.

The European
Semester is a 6-month cycle of economic and fiscal policy coordination, during which each country aligns its budgetary and economic policies with the objectives and rules agreed at the EU level with the Commission and the Council. 4. Before the Treaty of Nice larger countries had two commissioners. 5. The financial assistance support differs for euro and non-euro members. For the former, a permanent intergovernmental mechanism was created in December 2010: the European Stability Mechanism, which substituted the European Financial Stability Facility. The latter countries receive mutual assistance thanks to Balance-of-Payment assistance. In addition, the European Financial Stability Mechanism borrows money to all the countries. 6. Only since 2011, with the entry into force of the Six Pack, the budgetary surveillance mechanism has given more prominence to public debt also in the corrective arm through the operationalization of the debt criterion and the faster adjustment path towards the MTO for highly indebted countries. 7. The SGP has been under severe strain, despite lower noncompliance costs (Börzel et al. 2010), even before the sovereign debt crisis. Countries have violated the debt and deficit rules several times, have severely misreported their data (e.g. Greece) and employed creative accounting techniques (Alt, Lassen, and Wehner 2014;von Hagen 2010;von Hagen and Wolff 2006). 8. Baerg and Hallerberg (2016) use also a computer-based Levensthein distance metric to measure the frequency of editing of Commission's recommendations. dimension (see Crombez and Hix 2015). 14. Unlike Baerg and Hallerberg (2016), I do not consider the output gap, because it is associated with potential and trend economic activity and its estimate is subjected to heaviest revisions, especially when the economy is hit by unpredictable shocks. (Caudal et al. 2013 Appendix 1. From official documents to usable data: coding rules for the analysis of country-specific recommendations The official documents must be processed in order to become usable data for the empirical analysis. First of all, both the Commission's and the Council's documents are subdivided into major provisions, defined as an item (article, paragraph, subparagraph, sentence, indents or lists of elements) that provides substantive details about the stability or convergence programmes.
Major provisions are counted according to the following rules: 1) Articles and numbered paragraphs count as separate provisions; 2) Unnumbered paragraphs, subparagraphs, sentences, indents and lists of elements, even if numbered, count as separate provisions only if they require additional constraints on member states; 3) Unnumbered paragraphs, subparagraphs, sentences, indents and lists of elements, even if numbered, do not count as separate provisions if policy authority and constraints are related to the same.
Second, if the provisions clearly list the policies and reforms to be taken and implemented, in the form of indents, and upon invitation or request by the Council, I code them as 'recommending provisions'. Major recommending provisions are counted according to the following rules: 1) List of recommendations as clearly expressed in the Commission's and Council's documents (at the end of the document, after: 'HEREBY RECOMMENDS that' or '[the Member State] is invited to:' 2) In articles, paragraphs, subparagraphs, sentences, indents the Council 'invites', 'urges', 'recommends', 'encourages' the member state in doing/achieving a goal. 3) In articles, paragraphs, subparagraphs, sentences, the Council make explicit reference to a policy that 'needs to be pursued', 'should be pursued' or needs to be reached by a certain date (or not).
Provisions setting similar or vague recommendations are not counted. Third, I examine the propensity to as well as the direction of policy change between the Commission's recommendation and the Council's opinion on the macroeconomic programmes. I determine whether the Council's recommending provision can be easily associated with the wording of the Commission's text. If it is so, the provision is labelled as 'No Change' (NC). If the Council's recommending provision contains amendments to the Commission's text, then it can be associated as 'Strong' (S) or 'Weak' (W). Recommending major provisions that strengthen (S) the Commission's proposed recommendations are those: 1) Specifying that the Council imposes that member states shall take more and strengthening action (e.g. take more and detailed measures, report, provide information) than those recommended by the Commission.
2) Specifying that the Council imposes more detailed and shorter deadlines by which the member states shall take the requested actions than those recommended by the Commission. 3) Setting by the Council detailed rules for member state's actions (e.g. deficit reference values).
Recommending major provisions that weaken (W) the Commission's proposed recommendations are those: 1) Specifying that the Council imposes that member states shall take less and weakening actions (e.g. take measures, report, provide information) than those recommended by the Commission. 2) Specifying that the Council deletes or postpones the deadlines by which member states shall take the actions with respect to those recommended by the Commission. 3) Setting by the Council less detailed rules for member state's actions (e.g. deficit reference values) than those set by the Commission. 'As part of a wider effort to improve the efficiency of public administration, clarify competences at all levels of Government. Ensure better management of EU funds by taking decisive action to improve administrative capacity, transparency, evaluation and quality control at regional level, especially in southern regions.' 'As part of a wider effort to improve the efficiency of public administration, clarify competences at all levels of Government. Ensure better management of EU funds by taking decisive action to improve administrative capacity, transparency, evaluation and quality control at both at national and regional level, especially in southern regions.'

Strong
Cyprus 2011 'Abolish remaining obstacles to the establishment and free provision of services in sector-specific legislation by October 2011 in order to create more opportunities for growth and jobs in the services sector.' 'Abolish remaining obstacles to the establishment and free provision of services in sector-specific legislation by December 2011 in order to create more opportunities for growth and jobs in the services sector.' Weak distinguish whether these provisions either assess the national macroeconomic program or recommend policies and reforms. Moreover, Baerg and Hallerberg do not clarify how they have aggregated these estimates for each document and make replication hard. My procedure, on the contrary, focuses only on the specific recommending provisions, which are the core elements of the country-specific recommendations. I also explain how recommending provisions are analyzed and can be aggregated into measure of change and direction of change for each country-year document. This procedure reduces reliability problems and facilitate replication. Clearly, the differences between the two procedures are glaring and should be subject to further research.

Robustness checks
This section evaluates whether the empirical findings are robust in models if the dependent variables are operationalized differently. Instead of measuring the propensity of change in terms of absolute number of the amendments, Change is a dichotomous variable which takes value of 1 if at least one recommending provision of the csr is changed by the Council. As shown in Table 2A, I run a logit regression model and found that the relation between the independent variables and Change is actually reinforced. Only diffuse public support is not significant, but the relationship is still negative as on my expectations.
As one reviewer noted, this Change variable may suffer from validity problems because it does not capture the magnitude of change. For example, it does not distinguish whether the countryspecific recommendation has been subject to one or 10 amendments (as it is the case of Portugal in 2014). Nevertheless, it is reassuring to find the same results with a dichotomous dependent variable.
I also examine a more nuanced measure of the direction of change, which I label Severity. This new ordinal variable includes only the 265 country-specific recommendations being changed by the Council: Robust standard errors in parentheses ***p < 0.01, **p < 0.05, *p < 0.1 Models with country and year fixed effects In comparison to Council Severity, this new variable modifies only the category of country-specific recommendations that takes the value of 0, namely those containing the same amount of strengthened and weakened provisions. Table 3A reports the results from this new ordinal logit regression analysis. Country's political power and the economic crisis are stronger and more stable predictors of severity in the recommendations, especially in the pre-Six Pack period. Compliance records of the member states, instead, are no longer robust determinants. Although I think that the results from this corrected dataset are helpful in supporting my claim, I still believe that the analysis is most appropriate for the expanded set of country-specific recommendations. Robust standard errors in parentheses ***p < 0.01, **p < 0.05, *p < 0.1 Models with country and year fixed effects