Reconsidering the drivers of country-specific recommendations: The Commission's ideological preferences on wage policies

As part of the European Semester, the European Commission issues country-specific recommendations for all member states. I contribute to the literature on this political instrument, by considering the determinants of recommendations calling for greater wage moderation and enhanced cost competitiveness. For the most part, research on European economic governance has either understood the European Commission as a politicized and ‘ideological’ institution or as a de-politicized, technocratic actor. My analysis shows that the European Commission's ideological preferences on labour markets and wage bargaining institutions are more convincing predictors than explanations based on economic indicators. By testing a series of multilevel models, I find that irrespective of developments in competitiveness, countries with stronger social actors are more likely to be recipients of country-specific recommendations calling for wage restraint.


Introduction
The global financial crisis (GFC) and the Eurozone crisis have significantly enhanced the role of European Union (EU) institutions in the regulation of member states' labour markets. In the absence of currency devaluations, recommendations on labour market policies have acquired a particular importance for Eurozone states. In fact, labour market reforms leading to greater wage moderation are a key tool to improve countries' competitiveness indicators. Moreover, as policymakers have often understood the Eurozone crisis as a competitiveness crisis, wage and labour costs developments have become important macroeconomic variables to monitor in the context of a more integrated European policy cycle.
This article examines the European Commission's (henceforth 'Commission') preferences on labour market policies by focusing on one instrument in particular, which is country-specific recommendations (CSRs): a yearly set of member state-specific policy recommendations monitoring member states' policies in a wide variety of different areas. By doing so, I seek to make two contributions: one on the economic, political and institutional determinants of CSRs and the other on the preference formation of the Commission in the context of the new European economic governance and the increasingly important role this institution plays in member states' labour markets. Existing literature on the post-crisis' EU economic governance has identified two dominant strands of preference formation for the Commission: the first conceives the Commission as a de-politicized and technocratic institution implementing a rules-based system, the second instead considers the Commission as a politicized actor, whose recommendations are informed by a set of ideological preferences ( Van der Veer and Haverland, 2019). While the dichotomy between these research strands has animated investigations on the Commission's labour market policy preferences (Bokhorst, 2022;Schulten and Müller, 2015;Syrovatka, 2021b), an empirical investigation on the determinants of the Commission's recommendations on wage policies has thus far been missing. In this article, I show that the analysis of wage policies constitutes an important area of research for EU studies. As laid out in the Five Presidents' Report (Juncker et al., 2015), the need for a greater coordination on wage developments and competitiveness indicators has been found to be an important steppingstone in completing Europe's Economic and Monetary Union (EMU) and ensuring greater synchronicity between member states' labour markets. Moreover, recent policy initiatives by the Commission on adequate compensation for short-time work schemes, transparent working conditions and 'fair' minimum wages are indicative of the institution's growing interventionism in member states' labour markets and industrial relations. Finally, as in the years 2011-2019, calls for wage restraint and enhanced cost competitiveness have featured in 30% of all CSRs. This is clearly a policy area in which the Commission has been particularly active.
This article explores the question of whether CSRs on wage policies are more likely to occur in the presence of an economic downturn and when productivity indicators fall behind wage growth or whether, on the contrary, CSRs are more likely to be predicated upon institutional labour market characteristics. My findings show that the Commission is more likely to recommend reforms in wage policies in member states where social actors are strong and where wage-setting is highly centralized. More generally, this addresses the question of whether the policy recommendations of the Commission are informed by a technocratic logic, which ties recommendations to objective economic criteria or whether the Commission's preferences are associated with an ideological approach, in which country-specific recommendations may also be decoupled from the economic performance of countries. Therefore, while this article examines CSRs in a specific policy field, thematically, it seeks to contribute to the broader literature on the question of whether the Commission's preference formation is structured along technocratic or ideological lines.
Contrary to expectations, I find that increases to labour costs, which have often been seen by European policymakers as a proxy for competitiveness (Storm and Naastepad, 2015), do not seem to be associated with an increased likelihood of obtaining a CSR in the policy areas of wage restraint and cost competitiveness. This finding is robust across different operationalizations and modelling strategies. Variables, which measure the strength of social actors and labour market institutions, tend to be more convincing predictors instead. In fact, I show thatirrespective of the development of competitiveness indicatorscountries with centralized wage-setting institutions and high collective bargaining coverage rates are more likely to be recipients of CSRs calling for greater wage restraint and enhanced cost competitiveness. These results therefore suggest that the recommendations of the Commission seem to follow an ideological bias in the areas I examine. Furthermore, this empirical analysis also furthers our understanding on whether the Commission's policy outputs have undergone a process of socialization or liberalization. This is an increasingly debated aspect in EU politics, and one which has garnered a great deal of scholarly attention since the Juncker Commission's aspiration of guaranteeing Europe a 'triple-A social rating'. Empirically, this article's findings do indeed suggest that the focus of policy recommendations has shifted slightly in time. In fact, compared to the first CSRs, more recent recommendations have called for greater wage growth and for the need of fostering drivers of aggregate demand. However, I do not find evidence that the differences in time are large enough to be able to conclude that there are statistically significant differences between the Barroso and Juncker Commission. Moreover, given the exceptional nature of the European Semester's (henceforth 'Semester') post-pandemic policy cycle, it remains to be seen whether the current Commission will articulate future wage policy preferences in the direction of wage growth or wage restraint.

European integration, wage-setting regimes and the Eurozone crisis
While the equifinality, the extent and the degree of change in advanced capitalist democracies' industrial relations constitutes a lively topic of debate (Baccaro and Howell, 2017;Hassel and Palier, 2021), the direction in which change has occurred seems to be quite clear. In most countries, industrial relations and wage bargaining have been liberalized. During the financial crisis, EU-inspired or sanctioned labour market reforms have led to a further liberalization of labour relations (Schulten and Müller, 2015). Much of the impetus behind these reforms has hinged on the purported inflexibility of coordinated labour markets, where wage bargaining does not occur at the company level, but rather at the sectoral or industry level. Decentralized, firm-level wage-setting mechanisms were considered to be better equipped to respond to the vagaries of the market. In fact, at the height of the crisis, collective bargaining institutions were often seen as a hindrance rather than as an important factor in mediating the impact of economic transformations (Blyth, 2013).
The role played by the EU institutions in restructuring Europe's labour relations has elicited a great deal of attention. As noted by several commentators, during the GFC, the policy recommendations and interventions of the Commission were informed by a precise economic ideology, which saw in fiscal retrenchment, wage moderation and enhanced price competitiveness the way out of the crisis. These recommendations were in line with those held by the International Monetary Fund (Blanchard et al., 2014), which at the time often emphasized wage-setting decentralization and greater labour market deregulation for countries experiencing economic contractions. Furthermore, as the crisis was often understood by European policymakers as a competitiveness crisis (Johnston, 2016;Miró 2021a), with roots in the structural differences of member states' labour markets, policy recommendations frequently focused on ensuring that wage growth remained moderate. This was seen as fundamental in order to achieve established competitiveness targets. Consequently, wage developments and the harmonization of member states' labour markets were perceived as important mechanisms by which the imbalances of the Eurozone crisis could be redressed (Copeland and Daly, 2018;Syrovatka, 2021a). Moreover, during the crisis, the notion of competitiveness played a key role in justifying recommendations on labour market and wage-setting reforms. While competitiveness is a broad term that encompasses a wide range of policy areas, Miró (2021a) has compellingly illustrated that EU policymakers at the time of the crisis mostly understood this concept in terms of cost competitiveness. Consequently, the policy solutions to economic contractions prioritized lower wage growth and decreases in unit labour costs. In the context of the aftermath of the crisis, price competitiveness and wage development policies became central in the articulation of the Commission's policy recommendations (Syrovatka, 2021a).
At the height of the GFC, the Commission's policy preferences and recommendations not only focused on economic outputs, such as lowering labour costs or moderating wage demands, but also concerned the structure of the labour market. During the crisis, member states with poorly performing economies, high collective bargaining rates and coordinated wage-setting systems were particularly targeted as policy recommendations called for wage-setting decentralization and limits to the extension of collective bargaining agreements. The nature of these recommendations feeds into the question of whether different wage-setting regimes and labour market types are associated with heterogenous economic outcomes, which is a topic that has been subject to considerable debates in the EU's political economy literature (Johnston, 2016;Nölke, 2016). Höpner and Lutter (2018) have, for instance, found that decentralized wage-setting institutions are associated with higher wage inflation, whereas centralized and corporatist labour relations regimes tend to have lower wage inflation. The different outcomes that different constellations of wage-setting regimes give rise to has even led some scholars to question the compatibility of different wage-setting regimes and labour market institutions in a common currency area (Johnston and Regan, 2016). Moreover, as the establishment of a currency union effectively prevents member states from restoring competitiveness by devaluing their currencies, internal devaluation, in the form of wage restraint, constitutes an important and immediate tool to address competitiveness losses (Armingeon and Baccaro, 2012).
As this section has shown, the EU displays a wide variety of different political economies, whose successful integration into a currency union is often contingent upon the institutional framework of member states. Moreover, the crisis has shown that EU institutions preferred decentralized over centralized wage-setting institutions. The following section will focus more closely on how the policy preferences are articulated within the framework of the new European economic governance.

The European Semester and the role of the Commission
During the course of the crisis, the Commission's policy recommendations on labour market policies and wage-setting mechanisms were informed by a series of ideological tenets on how best to confront and recover from economic crises. Consequently, recommendations and policy interventions have been seen as biased in favor of austerity and a liberalization of the labour market (Blyth, 2013). Moreover, the preferences of Northern 'creditor' states held greater sway in the articulation of the EU's political economy than those of Southern, 'debtor' states, which bore the brunt of the crisis (Fabbrini, 2016). These beliefs have defined and structured the ideological contours of what has been termed the new European economic governance. More recently, however, some commentators have noted that the Commission's economic discourse has shifted away from the pervasive notion of austerity into a more 'social' and investment-oriented direction, as European policymakers have begun putting a greater emphasis on social rights. Miró (2021b) has also noted that after 2016, the Commission's recommendations showed evidence of encouraging a fiscal policy which was conducive to the promotion of policies designed to enhance aggregate demand. The commitment of the von der Leyen Commission to respect the autonomy of social partners and national wage-setting systems is also in contrast with the policy interventions that characterized the crisis' immediate aftermath (Rainone, 2020).
The important role that economic ideology plays in the labour market policy preferences of the Commission is in contrast with the 'traditional' understanding of the institution as a technocratic, de-politicized actor, implementing a rules-based governance system (Majone, 2002;Moravcsik, 1998). If the Commission's role is to be construed this way, then policy recommendations should be informed by 'evidence-based' criteria only (Zeitlin and Vanhercke, 2018), and should, consequently, be mostly untouched from the discretionary preferences, which an ideological affinity to an economic school of thought or another would presumably reveal. As the GFC has significantly enhanced the 'regulatory space' of the Commission (Bauer and Becker 2014) an understanding of what factors are associated with the Commission's labour market preferences is important. Before addressing the question of how the Commission's preference formation is articulated in policy outputs, this article will examine the increasingly important role that the Commission's labour market preferences play in the Semester (the EU's post-crisis annual cycle of policy coordination and socio-economic governance).
The GFC not only exposed several structural deficiencies of the EMU's institutional architecture, but also revealed the impact that the lack of macroeconomic coordination between member states had. The Semester, which was established in 2010, sought to address these shortcomings by strengthening coordination between member states in fiscal and macroeconomic matters. The crisis, however, also brought a range of ancillary socio-economic matters to the fore (de La Porte and Heins, 2016). In fact, the recommendations that the Commission issues as part of the Semester do not only include policies which can strengthen coordination in macroeconomic areas, but also overall country-specific plans, which can include policy areas which, formally, are not within the remit of EU competences. Although wage-setting is not part of the EU competences, the reforms package introduced by the EU in the 'Six Pack' regulations (2011), not only enables the Commission to monitor national wage policies, but also sees in wage policies an important lever in ensuring greater macroeconomic coordination between member states. The establishment of the Semester has also affected the governance of labour market policies. Pre-GFC policy coordination attempts between member states on labour market matters mostly relied on soft law, benchmarking and the Open Method of Coordination (OMC). However, with the establishment of the Semester, the coordination between member states became inscribed into a 'regime of hard-governed fiscal policy' (Zeilinger, 2021: 63). This in turn entailed that the EU's labour market governance acquired, as stated by Roland Erne, a 'vertical' dimension, whereby the articulation of the Commission's policy preferences plays an important role for member states (Erne, 2015).
The Commission outlines its priorities in Annual Growth Surveys, which are followed by country-specific recommendations, which can be amended or be approved by the Council. While the implementation of the recommendations varies from country to country and from year to year, CSRs provide an indication not only of the policies the EU would like member states to adopt on a national level, but also provide an insight as to what is needed to strengthen macroeconomic and fiscal coordination between member states. The codification of the Commission's preferences in the Semester has allowed the Commission's policy prerogatives to play a more important role in shaping member states' policy outputs (Mariotto, 2022). This trend has recently been accentuated: in fact, the COVID-19 crisis and the establishment of the post-pandemic Next Generation EU's Recovery and Resilience Facility has led to a greater institutionalization of the Semester (Vanhercke and Verdun, 2022).
The changes introduced by the new European economic governance regime have led to significant debates. A great deal of attention has, unsurprisingly, been directed to the evaluation of the contents of the policy recommendations issued by the Commission. Several investigations into the topic have thus focused on coding the policy directions of the CSRs. Zeitlin and Vanhercke (2018) have found evidence for a socialization of the European Semester, lending credence to the Juncker Commission's claim of aspiring for 'a more social Europe'. This is in contrast with the findings of other researchers (Copeland and Daly, 2018;Syrovatka, 2021a) who instead find that the policy recommendations of the Commission, as articulated in the Semester, veer in favor of market liberalization and a commodification of labour relations. More closely related to the topic investigated here, existing research on the Semester has focused on examining the determinants of CSRs in different policy areas. Azzopardi-Muscat et al. (2015) have found that the EU's recommendations on health policy are mostly tied to fiscal and macroeconomic metrics as opposed to topic-specific indicators such as the share of health spending. In their analysis of the drivers of pensions recommendations, Guidi and Guardiancich (2018) have expanded this approach by sub-dividing the CSRs on this policy area into a series of 'sub-CSRs'. Their findings are in contrast with Azzopardi-Muscat et al.'s (2015) analysis of health policy, as they conclude that the EU's recommendations on pension policy follow a mostly technocratic logic and are indeed tied to the relevant predictors in an intuitive way. Finally, Van der Veer and Haverland (2018) argue that the Commission is more likely to issue recommendations in countries where polarization on EU integration is higher.
CSRs do not simply constitute aspirational objectives by the Commission, but can also have tangible consequences for member states. First, from a legal standpoint, the issuing of recommendations is justified in different ways; those which find their legal bases in the Stability and Growth Pact (SGP) and the Macroeconomic Imbalance Procedure (MIP) are considered binding. While the Commission has thus far refrained from doing so, it is important to note that non-compliance to recommendations grounded in the SGP and MIP might lead the Commission to levy fines (Syrovatka, 2021a). Furthermore, recommendations favoring social retrenchment and wage restraint over social investment are more likely to be based in what Crespy and Vanheuverzwijn (2016) describe in their analysis as more 'solid legal foundations'. Second, even when not binding, CSRs can have an indirect impact on member states' reforms initiatives. CSRs on labour market flexibilization and wage-setting have, for example, had an important impact in the domestic political agenda of labour market reforms in Belgium, France and Italy (Bokhorst, 2022;Syrovatka, 2021b). Although the exact nature of the mechanisms through which member states comply with CSRs constitutes a burgeoning area of research (Guardiancich and Guidi, 2020), CSRs have often found to have had a significant impact on national reform programs. Finally, while the Treaty on the Functioning of the European Union (TFEU) recognizes the autonomy of social partners and national wage-setting mechanisms, the policy discourse espoused by the Commission can have clear implications on the policy demands articulated by European social actors (Erne, 2015). Jordan et al. (2021) have for instance found that the European Trade Union Conference's (ETUC) objectives are influenced by claims of a socialization of the Semester. CSRs can therefore have consequences for member states and for the social actors seeking to articulate the preferences of the workforce at the European level.

Theory and hypotheses
As briefly discussed before and following previous scholars (Guidi and Guardiancich, 2018; Van der Veer and Haverland, 2019), I expect the Commission's preference formation to be modulated either upon: (a) a de-politicized and technocratic logic, or (b) a politicized and discretionary/ideological approach. In the formulation of my hypotheses, I distinguish between these two approaches and, as illustrated below, I expect variables to be associated with the probability of a CSR in the policy areas of wage growth and cost competitiveness in different ways.

The Commission as a technocratic regulator
At the time of the crisis, developments in labour costs and cost competitiveness indicators played an important role as a justification for labour market reforms. This leads to the following hypothesis: H1: Increases to unit labour costs (ULCs) in the preceding year are associated with an increased probability of a CSR on wage developments.
As economic theory would predict, the higher labour costs are, the wider is the wedge between compensation and economic output and the less competitive an economy becomes. In other words, I do not only expect that CSRs on wage restraint and cost competitiveness are more likely in the event of higher labour costs, but that a technocratic Commission should be receptive to increases (decreases) to ULCs within a country, which should, in turn, be associated with an increased (decreased) likelihood of a CSR in the policy areas under analysis.
Furthermore, as wage costs are a key element which policymakers seek to address when confronting economic contractions (Blanchard et al., 2014), I hypothesize negative or sluggish GDP and employment growth to also influence the likelihood of a country being a CSR recipient in the fields of wage restraint and cost competitiveness.
H2: A decrease in the GDP growth rate increases the chance of a CSR in the policy areas of wage restraint and competitiveness.
H3: Unemployment growth increases the risk of a CSR in the policy areas of wage restraint and competitiveness.

The ideological preferences of the Commission
Considering the fact that the TFEU stipulates that wage-setting and the autonomy of social partners are a purely member states-specific competence, a technocratic Commission should be mostly informed by objective economic criteria when issuing recommendations on a member state's wage growth indicators. Provided that economic, employment and competitiveness indicators are positive, then different institutional labour market configurations should not matter in the articulation of the Commission's policy preferences. Should this, however, not be the case, then I expect the recommendations of the Commission to be informed by its discretionary preferences. This approach assumes that a series of institutional settings are better or worse equipped to deliver a more competitive economy.
H4: Countries displaying high degrees of wage-setting centralization have, ceteris paribus, a higher likelihood of receiving a CSR on their wage bargaining systems.
This hypothesis is, of course, in contrast with the rich and nuanced literature into the topic of wage bargaining centralization and wage restraint, which has found that different wage-setting constellations in the sheltered and traded sectors in the Eurozone's member states are associated with different outcomes in wage growth (Johnston and Regan, 2016). As wage-setting centralization is a (mostly) time-invariant variable, I estimate whether the Commission's likelihood of issuing a CSR on the topics I examine is also influenced by two time-varying variables, which seek to operationalize the industrial relations 'profile' of a country. These are the union density rate and the collective bargaining coverage rate. Considering that the TFEU guarantees the autonomy of social actors and wage-setting institutions and provided that competitiveness and economic indicators evolve in line with expectations, a purely technocratic Commission should not issue CSRs based on the development of collective bargaining coverage or union density rates.
H5: Higher union density rates and higher collective bargaining coverage rates are associated with an increased likelihood of obtaining a CSR in the areas of wage restraint and cost competitiveness.
Finally, I include the following control variables: the political leaning of the incumbent government as measured on the left-right and pro-anti EU axes of political competition, as well as a variable measuring the different Commissions. As discussed, the Juncker Commission's claim of aspiring for a more 'social' Europe has led some scholars to detect a socialization of the European Semester (Miró, 2021b;Zeitlin and Vanhercke, 2018). Following this reasoning, the Barroso II Commission should be associated with a decreased likelihood of issuing CSRs in the chosen policy areas compared to the Juncker Commission.

Dependent variable
CSRs do not only straddle different policy areas, but recommendations in one policy area, such as wage policies, can encompass different aspects. In this section, I explain the operationalization of the dependent variable in greater detail and expand on the type of recommendations included in the dataset.
The Commission's recommendations on the linkage between wage-setting and more competitive economies tend to focus on the need to ensure that wage growth does not exceed productivity growth (Schulten and Müller, 2015). In addition to this, recommendations on wage developments also include some more specific policy aspects, which leads me to include broader specification criteria. I include CSRs, which call for revisions to member states' indexation mechanisms, as well as recommendations suggesting further attention to the link between wage-setting and competitiveness in only specific sectors of the economy. Belgium has for instance often been the target of CSRs on the wage indexation mechanism. Cyprus and Croatia have instead frequently been recipients of recommendations that aim to ensure that wage growth does not exceed productivity growth in the public sector. Interestingly, almost 40% of CSRs on wage-setting and wage moderation concern recommendations on the minimum wage. While Ireland and Greece, which were subject to a program of economic adjustment, froze or decreased minimum wage rates, other European countries have not been immune from targeted recommendations in this area either. As this overview has illustrated, this article's focus is only on recommendations on wage policies inasmuch as these concern wage restraint and the competitiveness of a member state's economy. This leads me to exclude other topics relating to wage policies, which are not directly related to the role of wage-setting in maintaining competitiveness, such as income inequality and the gender wage gap (Table 1).
It is important to note that not all recommendations on wage policies and cost competitiveness concern wage moderation only: some CSRs have, in fact, also focused on stimulating wage growth and aggregate demand. This has concerned countries with significant current account surpluses such as Germany and the Netherlands. As shown in Figure 1, the number of CSRs calling for wage growth has increased in time. While the increase in recommendations arguing for greater wage growth can be construed as evidence for a socialization of the Semester, it is also important to remember, as noted by Jordan et al. (2021), that the progressive waning of CSRs calling for wage moderation might also be due to the prior implementation of the necessary reforms. Finally, due to the COVID-19 crisis, the 2020 Semester stands apart from the other years' policy cycles. In fact, the 2020 CSRs do not feature the same amount of macroeconomic recommendations as in other years. Recommendations on wage-related matters mostly concerned pandemic-related income support mechanisms and social protection programs (Rainone, 2020). Due to the exceptional circumstances of the 2020 Semester, it is difficult to make predictions on whether the more investment-oriented nature of the 2020 Semester's recommendations will continue to be the norm. This is the reason why I limit the timeline from 2011 to 2019. Note: CSR, country-specific recommendation.

Covariates
In this section, I present which economic, industrial relations and political covariates I expect to influence the likelihood of a CSR. I provide information on the operationalization of these variables as well as the respective sources. All covariates are lagged by one year as the recommendations of the European institutions to member states are based on the countries' performance on a series of indicators at time t −1 . Variables that are operationalized by utilizing the yearly growth rate are instead measured by using the percent change from time t −2 to time t −1 . Indicators on industrial relations are based on the widely used OECD/AIAS database on the Institutional Characteristics of Trade Unions, Wage-Setting, State Intervention and Social Pacts (OECD/AIAS, 2021). The main covariate used to estimate the strength of social actors and labour market institutions is that of wage-setting centralization. In fact, in countries with more centralized wage-setting institutions, social actors play a more important role as they have sufficient power to ensure that negotiated sectoral/ industry wage agreements are implemented across firms (Kenworthy, 2001). The wagesetting centralization variable (operationalized as Level in the OECD/AIAS database) therefore measures the predominant level at which wage bargaining occurs. The original five-level categorization of the variable has been recoded into three levels to ensure a more equal distribution of observations between the different categories. At the lowest level of wage bargaining centralization, wage bargaining is mostly decentralized and therefore occurs at the firm level. In countries displaying intermediate wage-setting centralization, wage-setting occurs on a sectoral as well as on a firm level. Finally, in countries with high wage bargaining centralization, wage-setting predominantly takes place at the central or sectoral level. Further information on the recoding can be found in the Online appendix. I also add two time-varying variables which capture the importance of social actors in a country: these are the union density rate and the adjusted collective bargaining coverage rate (OECD/AIAS, 2021). 2 The union density rate is defined as the number of union members as a share of total employees. The adjusted collective bargaining coverage rate instead measures the proportion of employees covered by collective bargaining agreements as a share of total employees.
The main predictor for CSRs in wage policies and cost competitiveness is, as discussed, unit labour costs. This is operationalized as the index of nominal unit labour costs, which measures the ratio of labour costs to labour productivity in the total economy. To verify the robustness of my findings, I utilize two additional variables to measure the labour costs in the business sector of the economy and in the public sector. These are operationalized as yearly growth rates in the Labour Cost Index (LCI) and are sourced from Eurostat.
The macroeconomic control variables of interest are yearly GDP growth rates and the yearly unemployment growth rates. Data for these variables are taken from AMECO (AMECO, 2022), the macro-economic database of the Commission's Directorate General for Economic and Financial Affairs (DG ECFIN). As noted by researchers investigating the empirical merits of the 'socialization of the Semester' thesis and touched upon in the previous section, it has been found that the Commission matters in the articulation of CSRs (Copeland and Daly, 2018). This motivates the addition of a dummy variable for the Commission, where the Barroso II Commission (2009 is coded as the reference category. Finally, I also add two control variables to capture the political inclination of member states' governments. I measure cabinet partisanship on two axes of competition: a leftright dimension and a pro and anti-EU dimension. The data is sourced from the ParlGov database (Döring and Manow, 2021), which is a dataset of parties, elections and cabinets in EU and OECD countries. Higher values on the left-right dimension denote right-wing parties and lower values indicate left-wing parties. For the pro and anti-EU dimension, higher values correspond to pro-EU governments and lower values instead denote cabinets with a more Eurosceptic inclination. As this dimension is measured on a party-by-party basis, I have used a weighted average in the form of a crossproduct to measure the ideological persuasion of ruling cabinets by weighing seat share and party positions on the left-right and pro-anti EU dimension on an annual basis. Implicitly, this weighting assumes that the left-right and pro-anti EU positions of ruling coalitions correspond to the seat-weighted policy positions of the different parties that make up the governments in power (Martin and Vanberg, 2014).

Method
As previous studies on the determinants of CSRs (Guidi and Guardiancich, 2018;Van der Veer and Haverland, 2018), this article also uses multilevel models with country-specific intercepts in its modelling strategy. This means that I expect that some countries are just more likely compared to others to be CSR-recipients. In this type of data structure, the country-year observations are clustered within a higher-level variable, that is countries. The underlying concern is that temporal dependence exists within groups and that this may violate the assumption of independent errors. The general model can thus be specified as follows: where β 0i is the group-specific intercept, which is allowed to vary by country i, β 1 x it denotes a time-varying covariate and β 2 z i a time-invariant covariate. While previous studies have mostly investigated the impact of time-varying covariates on the likelihood of a country being a CSR-recipient, the institutional variable measuring the strength of social actors and labour market institutions needs to be properly contextualized. In fact, by focusing only on variables, which vary by year and by country, scholars fail to estimate the effect of sticky and rarely changing labour market variables on the likelihood of CSRs and privilege an approach which assumes that recommendations are informed by the time-varying economic performance of a member state. These types of analyses are commonly conducted with fixed effects models. However, these models, which control for unobserved heterogeneity by only estimating within-group variation, lead to inefficient estimates in the case of rarely changing variables (Plümper and Troeger, 2007) and are unable to estimate time-constant variables by design. To offset these problems, I specify a mixed effects model: by including the within-group mean estimates of the independent variables as predictors, I can estimate the time-invariant variables as well (Bell et al., 2019). The aforementioned investigations into the determinants of CSRs have assumed that while intercepts can vary, the relationship between the independent and the dependent variable is homogeneous across countries. Applied to this case, this implies that once we account for the fact that some countries are more likely than others of being CSR-recipients, the impact of the main time-varying independent variable (unit labour costs) on the likelihood of CSRs should be invariant. In my analyses, I test empirically whether this relationship is verifiable in the field of CSRs on wage bargaining.
Finally, while I am fundamentally interested in assessing the impact that these covariates have over time, as a robustness check I also run cross-classified models which cluster observations in countries and years.

Empirical analysis
In this section, I present the findings that I obtain from the data analysis. First, I test for the need of specifying a multilevel model with country-specific intercepts. After showing that this is indeed necessary, I discuss the different modelling specifications I employed and analyze their results. If in a given country at a given year, a CSR on the topic of wage restraint or cost competitiveness has been issued then this is coded as a positive observation. Out of a total of 235 country-years (2011-2019), 71 country-years are coded as having obtained a CSR in the policy areas I examine. In my analysis, countries which were under an Economic Adjustment Program during the crisis are excluded, as these were not issued with country-specific recommendations. 5 Further summary statistics, model diagnostics and robustness checks are presented in the Online appendix.
The need for modelling the data with a multilevel model is confirmed by the results of a log likelihood test (χ 2 = 92.4 with 1 degree of freedom), which reveals that the variance between countries is statistically significant and ought to be considered. Indeed, in the Online appendix, I show that country-specific intercepts display significant variation. Therefore, in line with existing research (Guidi and Guardiancich, 2018;Van der Veer and Haverland, 2018), this modelling strategy includes country-specific intercepts.
In my initial set of simple multivariate models, I test whether labour costs or industrial relations variables constitute significant predictors for the likelihood of a country obtaining a CSR in the areas of wage restraint and cost competitiveness. As explained, a mostly technocratic Commission should, in light of the importance that labour costs play in the articulation of the EU institutions' cost competitiveness preferences, be mostly informed by developments in labour costs. Alternatively, a more ideologically inclined Commission might, irrespective of economic trends, favour decentralized wage-setting institutions over centralized wage-setting institutions, as the latter are perceived as better equipped to confront economic changes. As one can see from the coefficient plots displayed below, the degree of wage-setting centralization is a statistically significant predictor: higher degrees of wage-setting centralization are associated with a higher probability of a CSR in the examined areas. Interestingly, the results for lagged ULCs are close to zero and are not statistically significant. These results hold when I cluster observations within years as well as within countries and years. I also include two models which operationalize the industrial relations profile of a country with time-varying variables. This is done by utilizing the union density rate and the adjusted collective bargaining rates. I again find that labour costs are not a statistically significant predictor, but that the variables capturing the industrial relations profile of a country are. The picture that therefore emerges is of CSRs which are mainly informed by industrial relations variables rather than labour costs, which should, theoretically, be the most likely economic predictor for CSRs on cost competitiveness and wage restraint ( Figure 2). I expand on this first series of models, by adding a series of control variables: the unemployment rate, the GDP growth rate, the left-right and pro-anti EU partisanship indicators as well as a dummy variable denoting the Commission (Barroso II or Juncker). These control variables remain the same across the models, what changes is the operationalization of countries' industrial relations profile (wage-setting centralization, collective bargaining coverage and union density rates for models 1-2, 3-4 and 5-6 respectively) and the methodology used. As far as the methodology is concerned, in these models I consider what differences are revealed when specifying: (a) a model with country-specific intercepts and fixed slopes; (b) a model with country-specific intercepts and countryspecific slopes for unit labour costs. The model fit statistics (displayed in the Online appendix) indicate that the simpler, random intercepts-only models have a better model fit than the more complex, random slopes models which presuppose the existence of heterogeneous effects. In other words, the statistical analysis does not find any evidence that would indicate that the magnitude of the effect of labour costs on the likelihood of obtaining a CSR in the policy areas of interest varies by country.
While the statistical significance of the different variables varies across models, these analyses show that overall, unit labour costs, which should, theoretically, be the main economic, time-varying predictors for CSRs on wage restraint and competitiveness do not have a statistically significant relationship with the dependent variable. Although the direction of the effect of unit labour costs is mostly in line with the hypotheses (higher labour costs are associated with an increased likelihood of a CSR in wage restraint and competitiveness), the magnitude of the effect is very moderate and never achieves statistical significance. The analysis of the results for the control variables is in line with the expectations: increases in the unemployment rate and decreases in GDP growth rates are associated with an increased likelihood of a CSR. While the results show that Eurosceptic incumbents tend to be associated with a higher likelihood of obtaining a CSR in the fields of wage restraint, there is no evidence that the results indicate statistically significant effects. Moreover, while the Juncker Commission is associated with a lower likelihood of issuing CSRs, these results are once again not statistically significant ( Table 2).
The most significant predictor for CSRs calling for wage restraint and enhanced cost competitiveness can instead be found in the industrial relations profiles of countries. Therefore, while the existing literature on wage bargaining regimes in the EU has shown that different institutional arrangements lead to different wage growth outcomes and that foisting upon member states the same wage-setting and labour markets institutions is untenable (Scharpf, 2020), countries displaying highly centralized wage-setting systems and high collective bargaining coverage rates are more likely to be CSR recipients. Furthermore, I find that the direction of the effect of union density rates is in line with the hypothesis I formulated for a more ideologically driven Commission: higher union density rates are associated with an increased likelihood of CSRs. However, while the results point into the 'right' direction, I also note that they fail to be statistically significant. This might be because countries with high wage-setting coordination and influential social actors do not necessarily also have high union density rates.  Note: * p < 0.1; * * p < 0.05; * * * p < 0.01. CSR, country-specific recommendation.
Additional models presented in the Online appendix indicate that the results are robust across different modelling strategies and data measurement specifications. When operationalizing labour costs as the year-on-year LCI growth in the business economy and in the public sector, the findings indicate that the magnitude of the effect is once again very moderate and, in both instances, fails to achieve statistical significance. I also find that the effects presented above are robust when utilizing cross-classified models, where the data is clustered within countries as well as within years.
I further probe the relationship between the different variables and the likelihood of CSRs on wage restraint and cost competitiveness by estimating marginal effects. In Figure 3, I estimate the marginal effects of GDP and wage-setting centralization on the likelihood of CSRs in the areas I examine. While I find that higher GDP growth rates result into a decreased likelihood of CSRs, the effect of low economic growth is conditioned upon the type of wage-setting centralization regime. Compared to countries with low wage-setting centralization and weaker social actors, countries with high wagesetting centralization are more likely to be recipients of CSRs calling for greater wage restraint in the event of low GDP growth. The higher the degree of wage-setting centralization is, the more pronounced does this effect become.
When examining the link between wage-setting centralization, labour costs and the likelihood of CSRs on wage restraint and competitiveness (Figure 4), I find that the marginal effects of labour costs are moderated by the type of wage-setting regime. While increases to labour costs are associated with an increased likelihood of CSRs, the effect seems to be dependent on the type of wage-setting regime. Once again, countries with higher wage-setting centralization and thus stronger social actors are expected to be more likely recipients of CSRs. As shown in Figure 5, I obtain similar findings when changing the operationalization of the variable measuring the strength of social actors from the mostly time-invariant variable of wage-setting centralization to the time-varying variable measuring collective bargaining coverage rates.
A battery of statistical tests using different modelling specifications have failed to uncover any meaningful relationships between increases in labour costs and recommendations on wage growth and cost competitiveness. While for a technocratic Commission, labour costs should, theoretically, be the most significant predictor for CSRs in the policy areas of wage restraint and cost competitiveness, I do find that other economic variables are also associated with the increased likelihood of CSRs in the areas I examine. As recommendations concerning wage policies and competitiveness seem more likely to occur in the event of low economic growth, I do not find enough evidence to discount the argument that the Commission's preference formation is informed by a technocratic logic. However, I also note the robustness of the predictions on CSRs when using industrial relations indicators. Countries with higher degrees of wage-setting centralization and higher collective bargaining coverage rates are far more likely to be CSR recipients in the fields of wage policies and cost competitiveness than countries with more decentralized wage-setting institutions. This analysis therefore reveals that the Commission's recommendations on wage restraint and cost competitivenesstwo key macroeconomic and labour market policies in the context of the new European economic governanceare more likely to be informed by a set of ideological preferences. According to this analysis, the Commission, at least between 2011 and 2019, has found countries with weaker social actors and decentralized wage-setting systems to be better equipped in moderating wage growth.
Finally, in my descriptive and inferential analysis, I have found that recommendations calling for higher wage growth and stronger aggregate demand increased in time. The results also point to the fact that the Juncker Commission has been less likely to issue CSRs on wage restraint compared to its predecessor. However, given the wide confidence interval of this result and the conservative estimation strategy used, this analysis does not find enough evidence to conclude that this result is statistically significant. Once the Semester will have resumed its pre-pandemic policy cycle, it will be interesting to evaluate to what an extent the von der Leyen Commission's pledge to respect the autonomy of social actors and wage-setting institutions will be implemented.

Conclusion
In this article, I have tested different mixed models with varying random and fixed specifications and have thereby uncovered interesting and perhaps counter-intuitive relationships between a series of economic variables and CSRs. Substantively, this analysis has failed to show any sort of significant relationship between labour costs and CSRs in the policy fields I examine. While this would indicate that the Commission's preferences are perhaps not as closely informed by a technocratic logic as found by Guidi and Guardiancich (2018), I also note that CSRs in the policy areas of wage restraint and competitiveness are more likely to occur when GDP and employment growth is low. This means that I do not discount the possibility of a technocratic, de-politicized Commission entirely, but rather find more convincing empirical evidence suggesting that ideology is an important driver for the Commission's recommendations. As the numerous models presented in this article have shown, the Commission's likelihood of issuing a CSR in the policy areas I investigate is considerably higher if countries display highly centralized wage-setting systems and are characterized by high collective bargaining coverage rates. Thus, my findings are mostly in line with the analyses of scholars which have found evidence that within the framework of the new European economic governance the Commission has emerged as a politicized actor (Erne, 2015;Zeilinger, 2021).
Looking at the future, understanding the drivers behind the Commission's recommendations on wage growth and cost competitiveness is important for various reasons. First, because a greater monitoring of wage developments at the European level is not only recognized as important for the economic resilience of the Eurozone, but also because the Commission is playing an increasingly important role in the regulation of member states' labour markets. Recent initiatives such as the support and expansion of short-time work schemes or the proposal for the introduction of a directive on fair and adequate minimum wages across the EU are symptomatic of the Commission's increased willingness to intervene and promote greater coordination between member states' labour market policies (Syrovatka, 2021a). Second, while the mechanisms through which CSRs are implemented in national reforms programs constitute a promising avenue of further research (Guardiancich and Guidi, 2020;Mariotto, 2022), investigations into this area have shown that also non-binding CSRs on policy fields that are not within the remit of the EU's competences can have important consequences for member states' domestic policy agendas (Syrovatka, 2021b). Furthermore, a better understanding of the drivers of policy recommendations is not only salient for substantive reasons, but also because of the increasingly institutionalized role of CSRs and the Semester. Finally, while my analyses have shown that the number of recommendations calling for wage restraint have decreased in time, my evaluation of this topic has stopped at the year 2019. In terms of wage policies, the extraordinary 2020-2021 policy cycles of the Semester have mostly focused on income support mechanisms rather than on ensuring that wage growth keeps in line with productivity developments. As discussed, it is difficult to assess whether this marks the start of a long-term shift in the Commission's wage and labour market policy-related preferences. However, the current record high inflation rates experienced by the Eurozone might well give the Commission reasons to recalibrate its policy preferences on wage developments in the upcoming Semester.