Dismantling labour markets from above: the case of wage policy in the European Union

Abstract

Are countries reforming their labour market constrained by external political or financial institutions? This article analyses the supranational pressures impacting wage policy in European member states. Theoretically, it takes inspiration from the literature on international financial markets to analyse the key phases of a nascent multi-level policy cycle, where the European Commission plays a pivotal role. Empirically, it distinguishes between wage policy prescriptions (CSR database, 2011–2019, complemented with EUR-Lex) and reform events (LabRef database, 2008–2019) granting more or less protection to workers. Methodologically, it employs mixed regression models with country and year random intercepts finding out that: (a) the Commission, when recommending wage policy, unambiguously prefers reforms reducing workers’ protection, thus displaying preferences aligned with those of international financial institutions; (b) through external conditionality, the pressures stemming from such biased approach exert a tangible impact on national policymaking, ultimately resulting in weakened labour market institutions and reduced well-being of workers.

1. Introduction

The European sovereign debt crisis laid bare the inadequacies of several coordination mechanisms that underpinned the Economic and Monetary Union (EMU). The crisis unleashed a new set of supranational constraints to the formulation of national macro- and micro-level policies. If financial market pressures were expected (Mosley, 2004), increased supranational political pressures coming from EU institutions represented an additional channel eroding national policymaking autonomy.

A major innovation reshaping the European Union’s (EU) governance was the increased coordination in national policies through the European Semester and ancillary regulations (Bauer and Becker, 2014; Savage and Verdun, 2016). Through the Semester, the Commission and the Council determine the policy priorities for the EU, monitor national budgets and reform programmes, and issue recommendations. The main novelty, compared to previous coordination mechanisms, was that the Commission, in its country-specific recommendations (CSRs), began to pair fiscal policies and structural reforms, thereby recognizing their complementary nature (Laffan and Schlosser, 2016). There is agreement that ‘the Semester has provided the EU-level with greater access to the economic, fiscal and social policies of member states’ (D’Erman et al., 2019, p. 198).

One key area where supranational pressures have intensified is the labour market, including on unemployment and activation policies, employment protection, and, the focus of this article, on wage setting. Wage policy featured prominently in the Stand-by Arrangements (SBAs) or Economic Adjustment Programmes (EAPs) signed between, respectively, the International Monetary Fund (IMF) or the troika, that is, the Commission, IMF, and the European Central Bank (ECB), and the seven member states that asked for financial assistance during the crisis. Furthermore, Italy and Spain were informally targeted by the ECB through the famous Trichet letters, while the Financial Sector Adjustment Programme (FSAP) for Spain explicitly referred to the fulfilment of CSRs related to the labour market. Despite having only limited competence according to the Treaties, within the European Semester, the Commission issued CSRs containing wage setting prescriptions to 16 of the EU-28 member states during the 2011–2019 period (countries under EAPs are temporarily excluded), mostly endorsed by the Council without modifications (see Supplementary Appendix Table A.1). In sum, only seven EU countries were spared (Austria, Czechia, Denmark, Poland, Slovakia, Sweden and, until it was a member state, the UK).

The literature dealing with wage setting in the EU has established that a new supranational institutional framework exerts greater pressure on member states. Scholarship is also in agreement on the direction of change: initially EU institutions pushed for wage moderation and budget restraint, relaxing their intransigence later on (Cova, 2022; Syrovatka, 2022). What is, however, missing, except for qualitative mapping exercises (Glassner et al., 2011; Clauwaert and Schömann, 2012; Vaughan-Whitehead, 2013; Marginson and Welz, 2015; Schulten and Müller, 2015; Pedersini and Leonardi, 2018), is a systematic assessment of the factors that influence a country’s propensity to reform.

The question regarding implementation is far from trivial. As previous research has shown, the CSRs often carry the potential to reduce the protection of workers, frequently transcending ‘objective’ competitiveness concerns. We are, hence, interested in showing how much reforms followed these recommendations. Assessing the external pressures exerted on domestic decision-makers may show whether a supranationalization of policymaking that is prone to systematic bias has taken place or not.

This article brings the research agenda significantly forward, adopting a comprehensive approach that analyses two key phases of the multi-level policy cycle on wage setting: the formulation of CSRs and the following implementation of reforms. The theoretical framework developed by Mosley (2003) and integrated by Guardiancich and Guidi (2022) provides the causal model of the relationship between international financial markets and EU conditionality, on the one hand, and national policymaking, on the other.

Instrumental for the analysis, we constructed an integrated dataset (see Supplementary Appendix Tables A.1 to A.4 for details) of:

• wage policy prescriptions, parts of either EPAs/SBAs (retrieved from EUR-Lex) or CSRs adopted by the Council, contained in the CSR Database, published by the DG for Economic and Financial Affairs, for the years 2011–2019. Recommendations range from extremely detailed (e.g. explicitly asking for changes in minimum wages) to generic (e.g. align wages with productivity), often but not always related to the overall macroeconomic situation;

• reform events, using the LabRef—Labour Market Reform Database, managed by the Commission in cooperation with the Employment Committee, for the years 2008–2019. Reform events relate only to the policy output, while the macroeconomic consequences (the outcome) are not captured in LabRef.

In view of assessing the policy direction, the recommendations and events were coded according to their potential to enhance or reduce the well-being and/or protection of workers.

The findings fall into two categories. First, with regards to the formulation of CSRs, the article confirms that certain institutional characteristics of the labour market, which are chiefly pro-labour, such as greater centralization of wage bargaining or having a statutory minimum wage, increase the likelihood of being recommended to reduce workers’ protection. Additionally, we find that ‘objective’ macroeconomic indicators are not entirely shunned, thus confirming that wage policy is instrumental in reducing internal and external imbalances, both fiscal and in terms of price competitiveness (EPSCO Council, 2011). Second, domestic implementation follows an external conditionality logic. Financial market pressures, deriving from increasing bond yields, as well as political pressures exerted through CSRs, trigger reforms in the desired direction. Of course, the state of the economy and institutions matter as well: GDP growth and lower current account deficits lessen the pressure on workers’ protection, while greater coordination of wage bargaining, a good proxy for institutional capacity, leads to more decisive action—in line with some industrial relations literature (Höpner and Lutter, 2018). Overall, the anti-labour drive in wage policy was highest under the Barroso II Commission, that is, during the European sovereign debt crisis, tapering off under Juncker.

In sum, we show that the Commission is not entirely objective when recommending wage policy and that such approach, at least partly biased against workers’ protection, exerts a tangible impact on national policymaking.

2. The ‘supranationalization’ of wage setting

A literature review on the supranational pressures on micro-level policies, such as wage setting, has to consider two viewpoints: (a) what the institutional reasons for the loss of national control are, and (b) how supranational pressures influence national policies. While researchers agree on the extent of institutional change and, often, also on its direction, the attempts at answering the second question are so far both theoretically incomplete and methodologically weak.

Whether it is referred to as the EU’s ‘New Economic Governance’ (Erne et al., 2024) or a ‘New European Labour Policy regime’ (Syrovatka, 2022), the impact of the European Semester cannot be underestimated. The mechanism aggregated all EU interventions related to the Stability and Growth Pact (SGP), to the Macroeconomic Imbalance Procedure (introduced in 2011) and to EU’s Europe 2020 strategy (Jordan et al., 2021).

The conditionality attached to several instruments (e.g. CSRs act as potent guidelines, being supported—depending on the legal base—by potential sanctioning mechanisms and, in the most extreme cases, the troika’s ‘shadow’) has been increased, intensifying EU’s leverage on socioeconomic policies, ranging from healthcare to pensions and the labour market (Azzopardi-Muscat et al., 2015; Schulten and Müller, 2015; Bokhorst, 2022; Maccarrone et al., 2023). Thus, the labour policy instruments of the EU have been strengthened politically, strategically and administratively (Syrovatka, 2022), entailing a leap (Maccarrone et al., 2023) ‘towards vertical integration, where workers and unions are constrained by direct interventions by a supranational authority’, which can ‘request policy changes in key areas of labour politics such as wage-setting mechanisms’.

Having established the EU’s increased clout, a fierce debate has erupted regarding both the Semester’s social credentials and its effectiveness. While some authors stress its progressive socialization (Zeitlin and Vanhercke, 2018; Vesan et al., 2021), others show that European social policy forcefully supports market development and functioning (Copeland and Daly, 2018; Haas et al., 2020; Jordan et al., 2021). Part of the exchange focussed on the analytical and methodological choices that inform each study and influence the assessment of the Semester’s orientation and impact (e.g., how to measure CSRs, to which extent their different legal bases affect conditionality, whether to include the overwhelmingly market-oriented prescriptions under EAPs; for a detailed discussion, see Erne et al., 2024). While, in line with the need to be as unambiguous as possible, individual recommendations have to be carefully scrutinized (especially in their policy context; Supplementary Appendix Table A.1 includes not only the text of policy prescriptions but also the relevant recitals in CSRs and evaluations in Country Reports), most authors agree that, at least initially, the overall policy orientation in wage setting was clear.

Armingeon and Baccaro (2012) posit that, in the aftermath of the crisis, the primary objective of the Commission was internal devaluation. Scholz-Alvarado (2022) claims that EMU policymaking followed the ‘logic of competitiveness adjustment’, not dissimilar from Syrovatka’s (2022) ‘economic resilience’, aimed at increasing productivity and containing labour costs, including in the public sector (see Vaughan-Whitehead, 2013; Di Carlo, 2023). Schulten and Müller (2015, p. 331) highlight a ‘one-sided focus on fiscal austerity and cost competitiveness’ that considers downward wage flexibility to be the main macroeconomic adjustment mechanism. Even more, Cova (2022) shows that CSRs are biased against centralized collective bargaining institutions and stronger social actors, which dovetails with Bongelli’s (2018) insight that internal devaluation is also recommended to countries where wages lag behind productivity growth.

As the crisis waned, the EU’s focus shifted away from austerity, after, however, leaving deep political scars, especially in the Union’s periphery. The Juncker Commission attempted to shore up support for the European project by aiming at a ‘triple-A social rating’ for the bloc, an aspiration later shared by his successor Ursula von der Leien. Hence, not only socially-oriented initiatives have mushroomed—for example, the European Pillar of Social Rights (EPSR) or the Minimum Wage Directive (Schulten and Müller, 2021; Vesan et al., 2021; Corti, 2022)—but also the CSRs have changed direction, calling for wage growth supporting aggregate demand. Even so, scholars are sceptical about the genuinity of the Commission’s U-turn. Whereas Cova (2022) speculates that fewer recommendations on wage restraint are due to cuts having been already carried out, Syrovatka (2022) recalls that past socially-oriented EU initiatives were often replaced with market-oriented ones.

Although outside the scope of this article, the reasons for the Commission’s shifting preferences are related to the coordination of prices in heterogeneous European labour markets required for the smooth functioning of the EMU. Of the two antithetical solutions, the first being the coordination of national corporatist institutions and the second being their dismantling and liberalization to unleash the coordinating power of market competition (see Scharpf, 2010), the latter was firmly entrenched in the discourse of EU institutions (in particular, of the ECB) from the very inception of the EMU. The panic following the sovereign debt crisis only reinforced this view. The change in stance after 2014 pertains to both issues of legitimacy, as EU institutions faced a popular backlash against austerity, and of governability, after the euro area drifted into deflation, putting under question the wisdom of the previously-advocated structural reforms (see Braun et al., 2022).

In sum, the existence of supranational pressures and, at least partly, their direction are not called into question. The elephant in the room is to what extent they impacted domestic decision-making, a question that has been so far only partly addressed.

Four factors appear to influence the extent of changes to wage-setting mechanisms since 2008 (Glassner and Keune, 2010; Marginson and Welz, 2015; see also Pedersini and Leonardi, 2018): (a) the depth and length of the economic downturn—the more persistent the initial shock, the bigger the expected change; (b) changing political opportunity structures, whereby supranational pressures make reforms more likely; (c) industrial relations institutions—wages are expected to be more responsive where collective bargaining is already decentralized; (d) social partners’ strategies, aimed at securing their involvement in policymaking, in order to achieve negotiated change. Several studies on wage setting in EU member states lend plausibility to these explanations (Glassner et al., 2011; Clauwaert and Schömann, 2012; Marginson and Welz, 2015; Schulten and Müller, 2015). With regards to the broader impact of the Semester, scholars have mainly analysed the Commission’s evaluations of past CSRs, testing which factors drive various policies’ implementation rates across the member states (Darvas and Leandro, 2015; Gros and Alcidi, 2015; Al-Kadi and Clauwaert, 2019; Mariotto, 2022; Efstathiou and Wolff, 2023).

Both types of studies may offer some room for improvement, especially by complementing each other. The findings of qualitative research are necessarily based on a limited number of cases, which present challenges in their generalizability. To this end, quantitative studies embracing all member states may corroborate their external validity by providing systematic explanations of the causes and effects of supranational pressures. Yet, even these studies have lacunae. In fact, they do not incorporate ‘the output side, namely the domestic political process and the likelihood of the Semester to shape national reforms’ (Haas et al., 2020, p. 339). These studies are, in reality, just scrutinizing the evaluations of the output provided by the Commission. Such an approach has three shortcomings. First, many reforms happen in the absence of recommendations. Second, it must be demonstrated (rather than assumed) that evaluations are correlated with reforms. Third, even if there is correlation, such measurement tells us nothing about the content of domestic interventions. Although most research (Verdun and Zeitlin, 2018; Zeitlin and Vanhercke, 2018) defines the Semester as an annual policy coordination cycle, the quantitative analyses do not treat it as such. In fact, nowhere are policy inputs (the CSRs) systematically linked with the outputs (domestic reform events). We attempt to fill in this gap.

3. Recommendations and reforms compared

Supplementary Appendix Tables 1 and A.3 show a breakdown of wage policy prescriptions, retrieved from the CSR Database and, manually, from the EAPs and SBAs (retrieved from EUR-Lex), plus policy reform events collected from the LabRef Database.

The two datasets employ a broad definition of wage-setting reforms/prescriptions, including measures that only indirectly influence wage policy, such as the representativeness criteria for the social partners or some individual rights of employees, for example the right to strike.

With regards to CSRs, 89 individual recommendations (sub-CSRs or sub-recommendations) related to wages and wage setting have been released between 2011 and 2019 (see Supplementary Appendix Table A.1), of which a sizeable number were broken down into distinct parts (e.g. requiring not only wage developments but also wage indexation to reflect productivity), thereby yielding a total of 110 individual policy prescriptions. As for the LabRef Database, 378 distinct reform events took place in the period of analysis, to which we added three regarding minimum wages (two for Italy and one for Slovenia) plus one on public employees for Spain, as they were only subsequently mentioned in LabRef. Unlike CSRs, we decided against breaking down each reform event into its constituent parts, because of the heterogeneity in the details provided. This has two implications.

First, some categories are less precise than in other authors’ works. While Marginson and Welz (2015, p. 442) distinguish between main levels of bargaining and the linkages between levels (hierarchy, opening and opt-out clauses, bargaining competence), here we group similar recommendations and reforms under the label ‘collective bargaining centralization’, following, hence, Schulten and Müller (2015). Second, a limited number of reform events contain multiple categories that have not been separated. The 2009 tripartite agreement in Italy contains provisions on decentralization and collective agreements’ expiry. We counted it as a single reform event.

As the LabRef grouping of reforms into five clusters was inconsistently applied across the dataset, we decided to break it down and thus homogenize the different datasets as shown in Table 1.

3.1 Policy direction general trends

The literature on the Semester has widely debated the measurement of policy direction. When analysing pensions, Guidi and Guardiancich (2018) and Guardiancich and Guidi (2022) distinguish between reforms promoting fiscal sustainability versus social adequacy. Crespy and Vanheuverzwijn (2019) and Vesan et al. (2021) differentiate between social retrenchment, investment and protection; Copeland and Daly (2018) between market-making, market-correcting, and mixed policy orientations. D’Erman et al. (2021) discerns between recommendations that strengthen or weaken the protection of workers. Copeland (2019), Erne et al. (2024) and Stan and Erne (2023) place recommendations on the commodification versus decommodification continuum (relative to the status quo).

Given the pros and cons of the different classifications, we decided to mark each recommendation and reform event as ‘pro-labour’, ‘anti-labour’ or ‘neutral’, depending on whether it increases, decreases or has mixed, limited or undefined effect on the well-being of workers and/or their protection from the vagaries of the labour market, contingent on the immediate labour market conditions and macroeconomic characteristics, relative to the status quo (see Supplementary Appendix Table A.3). The classification is inspired by D’Erman et al. (2021) and Erne et al. (2024). Yet, with regards to the latter, it avoids the pitfalls of placing policies on the commodification-decommodification continuum, when commodifying policies are socially progressive and vice-versa. Recommending that wages grow in line with productivity—basically the same wording was used in sub-CSR 3.3 for Germany in 2012 and in sub-CSR 2.1 for Belgium in 2016 (Supplementary Appendix Table A.1)—is an unambiguously ‘qualitatively’ commodifying recommendation, but has completely different ‘quantitative’ consequences for workers depending on whether wages are considered depressed (decommodifying) or vice-versa inflated (commodifying) (see Erne et al., 2024, p. 61). With regards to the former, instead, our study breaks down wage policy into detailed categories and is more context specific. As intersubjectivity was potentially an issue, the coding has been carried out by two independent coders, resulting in a weighted Cohen’s κ coefficient of 0.81. Supplementary Appendix Table A.4 shows the net sum of pro- and anti-labour prescriptions and reforms in each of the EU-28 member states during the period 2008–2019. The bar charts in Figure 1 display both pro- and anti-labour recommendations/reforms by country and year. Clear trends are discernible.

First, the timing of prescriptions and reforms conforms to expected patterns. Starting with CSRs, these tended to markedly run against workers’ protection at the height of the crisis. After the Barroso II Commission left office, the number of recommendations abated and their net sum shifted in favour of labour already by 2018. The alternation of reform events was even more pronounced. In 2008, pro-labour reforms still prevailed, while during the peak of the crisis (2009–2013) the opposite was true. From 2014 onwards, however, expansionary reforms started again to outnumber restrictive ones.

As for individual countries, Estonia, Germany and the Netherlands were the only member states receiving CSRs favourable to workers. Whereas the gender pay gap was targeted in Estonia, Germany and the Netherlands were advised to increase real wage growth to prop up domestic demand. Looking at reform patterns, there is little contradiction with the CSRs at the pro-labour end of the spectrum. While Germany and the Netherlands did what the Commission asked for, Austria, Poland and Sweden never received any labour market recommendations, implying that their policymakers had freer hands.

Of the member states advised to act against workers’ protection, the five countries that received greater attention from the Commission were Belgium, Luxemburg, France, Italy and Portugal. That they only partly overlap with the five countries that enacted the most restrictive reforms, that is, Greece, Cyprus, Spain, Portugal and France, is unsurprising. Supplementary Appendix Table A.2 shows that the countries that signed an EAP or SBA were subject to more stringent conditionality and were basically forced to overhaul their labour market institutions in 2009–2013.

3.2 Policy direction per category

Looking at individual categories, Supplementary Appendix Tables A.1–A.4 confirm that, from a supranational point of view, public sector wages are simultaneously treated as a competitiveness and as a fiscal problem (Di Carlo, 2023). If the impact of public sector wages on budgetary expenditures is clear, their relation to competitiveness is less so. Arguably, wage restraint in non-traded sectors has little (or even negative) impact on economic growth, as opposed to wage moderation in traded sectors (Johnston, 2021). Such ‘combined’ view was espoused by the fiscal provisions contained in EAPs and SBAs that recommended cuts in public wage and employment levels. Instead, the five CSRs explicitly aimed at public sector wage setting, for Croatia and Portugal (three CSRs implicitly also targeted public wages for Cyprus), only asked for lesser labour market segmentation. That public sector reforms depend on economic growth and budgetary leeway seems to also apply to individual member states: as fiscal conditions deteriorated in 2008–2013, the public sector was chastised through cuts in 16 EU countries (see Vaughan-Whitehead, 2013), while as many loosened the public purse during the subsequent 2014–2019 period.

As mentioned above, different reform dimensions, which often apply to multi-employer bargaining regimes, go under the ‘collective bargaining centralization’ label. In this respect, it is worth recalling that two important variables used to classify wage-setting systems are the degree of centralization, that is, the ‘level(s) at which wages are bargained or set’, and of coordination, that is, the ‘degree of intentional harmonization observed in the wage-setting process’ (Kenworthy, 2001, p. 59, 76). Compared to centralization, coordination is a difficult-to-measure concept. Accordingly, at the EU level, (de)centralization is considered as a key factor enabling labour markets to clear. An example is the 2016 Country Report for Estonia, which praises its decentralized wage-setting system as the bloc’s most efficient (Bongelli, 2018).

Even though the predominant level at which wage bargaining takes place in EU countries has undergone major changes since 2008 only in Bulgaria, Greece, Hungary, Ireland and Romania (see Supplementary Appendix Table A.5), a tendency towards decentralization is visible (OECD and AIAS-HSI, 2023). Throughout the period, as many as 16 EU-28 member states, especially those that joined the EU after 2004, had either single-employer bargaining regimes or mixed ones, where wage bargaining alternates between the sector and enterprise levels.

Supplementary Appendix Table A.4 shows that eurozone countries with high public debt, which resort to multi-employer bargaining, that is, Belgium, Greece, Italy, Portugal, Spain, plus hard-hit Finland, were recommended to decentralize to allow firm- or local-level bargaining. The 2011 SBA for Romania explicitly demanded that its new Social Dialogue Code abolish the national collective bargaining system. Several reforms affected the vertical coordination between bargaining levels, for example, their ordering, the existence of opening and opt-out clauses and of the favourability principle (see Marginson and Welz, 2015; Baccaro and Howell, 2017; Pedersini and Leonardi, 2018). The LabRef dataset confirms that all member states that received recommendations, bar Belgium, decentralized their bargaining systems. In addition, reforms reducing workers’ protection were introduced in France (primacy of company-level agreements, including on bonuses), Hungary (opt-out clauses and the right to conclude collective agreements for works councils) and Ireland (unconstitutionality of employment regulation orders setting sectoral minimum pay rates). Finally, during the 2014–2019 period, some centralization, for example, facilitating sectoral collective bargaining, took place, including in single-employer and mixed regimes, such as Ireland, Latvia, Luxembourg, Poland and Romania.

Two other aspects of collective bargaining—the extension/termination of collective agreements and the representativeness of the social partners—are only mentioned in the EAP for Portugal and the 2011 SBA for Romania. Yet, they were the object of reforms in 17 member states. The extension of collective agreements to non-signatory parties has been curtailed in Greece and Portugal, changed several times in Ireland and Slovakia, and expanded in Bulgaria and Germany, among others. The continuation of collective agreements beyond expiry has been limited in Croatia, Greece, Portugal and Spain. Additionally, the criteria for the participation of the social partners to various levels of collective bargaining were relaxed.

The relative majority of all CSRs focussed on the need that wage growth does not exceed productivity growth, whereby public sector wages were explicitly targeted in Cyprus and Croatia (Cova, 2022). Also the EAPs for Cyprus, Greece and Portugal, as well as the 2009 SBA with Latvia, recommended wage moderation. Finally, Germany and the Netherlands, both of which had sizable current account surpluses, were recommended to relax wage restraint. With regards to reforms, wage developments closely followed economic growth: if in the aftermath of the crisis, wage moderation was (almost) the only game in town, the reverse being true between 2014 and 2019, when only Belgium, Finland and Luxembourg enacted reforms penalizing workers.

A separate but tightly connected issue was the indexation of nominal wages to consumer prices. The ECB (2008) argued that such policy should be avoided because it might amplify inflationary shocks, leading to wage-price spirals. Just before the sovereign debt crisis, seven member states employed various indexation mechanisms: both public and private sector wages were indexed in Belgium, Cyprus and Luxembourg; private wages only in Spain; public and minimum wages in Malta and Slovenia; minimum wages in France. Countries with indexation guidelines in place were Finland, Greece and Italy. Supplementary Appendix Table A.4 fully reflects ECB’s views. All countries with automatic indexation, bar France, were issued at least one CSR. Cyprus also received recommendations on this within the EAP. Of the countries receiving these recommendations, only Malta and Slovenia did not carry out substantial reforms, while some adjustments lowering workers’ protection were introduced also in France and Greece.

With regards to statutory minima, we distinguish between minimum wage levels and the minimum wage-setting mechanism. Not all member states have statutory minimum wages. Austria, Denmark, Finland, Italy and Sweden often set sectoral minimum remuneration through collective agreements. Elsewhere the minimum wage-setting machinery takes on different forms with varying degrees of involvement of the social partners. The tendency has been to introduce a statutory minimum wage when collective agreement coverage falls under a certain level, as Germany did in 2015 (Mabbett, 2016). Cyprus followed in 2023.

The supranational attitude towards minimum wages has shifted over time. In the aftermath of the crisis, minimum wages were seen as hindering competitiveness and flexibility in the labour market (Johnston and Regan, 2016), given their spill-over effects across the wage distribution. Consequently, all countries under an EAP received recommendations on minimum wage levels. For example, the 2011 EAP for Ireland requested a 1 Euro reduction per hour (it was soon reversed). CSRs were more modest: due to previous minimum wage hikes, only France and Slovenia were targeted. As for the wage-setting mechanism, the 2010 EAP for Greece recommended the introduction of sub-minima for groups at risk, such as the young and long-term unemployed, while CSRs were issued to Bulgaria and Romania, aimed at establishing transparent and objective wage-setting criteria (in Romania the CSR was triggered by previous minimum wage increases). After 2017, however, recommendations regarding statutory minima basically disappeared. Moreover, a paradigmatic shift happened with the Minimum Wage Directive (Schulten and Müller, 2021), whose main goal was to reduce in-work poverty across the member states.

On the reform front, the difference between the two periods could not be starker. With regards to minimum wage levels, during 2008–2013, wage freezes prevailed over the rare hikes, especially in member states under financial strain. As growth resumed, 16 member states increased statutory minima. Only Belgium enacted restrictive reforms in 2014–2019, lowering the minimum gross salary for younger employees. With regards to the mechanism itself, little retrenchment took place. The biggest change happened in Greece, where a 2012 law determined that, after the end of the EAP, minimum wages would be set by the government following consultations with the social partners rather than through the National General Collective Employment Agreement.

Finally, while not denying their importance, a few issues tangential to wage setting, such as the gender pay gap (CSRs issued to Estonia), other employee rights (strikes plus information and consultation) and labour disputes (the 2010 EAP for Greece recommended to eliminate asymmetries in arbitration), were expunged from the statistical analysis.

4. Hypotheses

How can this interplay between national policies and supranational pressures be framed theoretically? The relationship between global economic integration and domestic politics is one of the most researched topics in international political economy (see e.g. Frieden, 1991; Oatley, 1999; Berger, 2000). Within this literature, the EMU has attracted considerable attention because of its unique combination of deep economic and political integration (Iversen et al., 2016; Frieden and Walter, 2017). Particularly relevant to this article is the impact of financial, legal-political and macroeconomic pressures on domestic reform propensity (see Efstathiou and Wolff, 2023, for an overview). Hence, our theoretical framework builds on Mosley (2003) and Guardiancich and Guidi (2022), who developed a causal model of the impact of supranational pressures exerted by international financial markets and the EU and/or IMF on national macro- and micro-policy, in our case wage-setting reform events, subject to domestic constraints.

Starting with supranational financial pressures, in a world of high capital mobility, financial actors exert their pressure by evaluating the risk of default of a certain country. Among the observed indicators, the budget deficit and debt are key (in the EU, these are defined according to the Maastricht criteria, see Mosley, 2004). When a state’s capacity to regularly service its debt is impaired, investors respond by selling the country’s government bonds, which in turn increases the interest rate that the country must grant to refinance its debt. Concerning supranational political pressures, the Commission and other institutions, such as the ECB and the IMF (i.e. the conveyors of external policy preferences) are also concerned with macroeconomic imbalances, as they may cause instability and spread through different channels to other countries, as well as with national competitiveness. Depending on how severe the crisis is, the supranational actors may recommend reforms and ensure they are enforced through several conditionality mechanisms.

National policymakers assess such supranational signals while being faced with the trade-off between the pressure to introduce macro- and micro-policy adjustments and their electoral costs. Hence, the direction and intensity of their reform effort depend on the interaction between financial and political pressures on the one hand, and the characteristics of the national system on the other—in the present case, not only its political and economic institutions but also the existing wage-setting mechanisms. The policymakers’ actions are then evaluated by both financial actors and supranational institutions, which may as a consequence increase or reduce the exerted pressure.

In sum, the theoretical framework encompasses three phases of an emerging multi-level policy cycle: the formulation of recommendations by external actors, the domestic implementation of reforms and their evaluation, again at the level of the EU/IMF (see Guardiancich et al., 2022). The focus of this article is on the first two stages of this cycle: supranational recommendations and domestic reforms.

With regard to the first, we integrate the analysis of Cova (2022), who distinguishes between a technocratic and an ideological approach of the Commission when recommending wage moderation (in line with the qualitative studies that emphasize internal devaluation, see, e.g. Syrovatka, 2022). We formulate two complementary hypotheses that posit a relationship between a country’s macroeconomic and policy indicators and the number of recommendations it receives. First, we want to test to what extent the Commission bases its prescriptions on objective statistics that summarize the functioning of the labour market, such as unit labour costs and the employment rate, or indicate incipient macroeconomic imbalances, for example, a growing public wage bill and current account deficits. These indicators are regularly mentioned in the Commission’s documents and employed as supporting evidence to push for wage moderation and labour market flexibility. Although lowering workers’ protection is a questionable approach to solving competitiveness-related and macroeconomic issues, the silver lining is that a country underperforming in these domains can be said to have problems to address. Hence, we hypothesize that:

H_1a: The Commission issues recommendations on wage policy based on relevant competitiveness indicators, labour market statistics and economic fundamentals.

Second, as shown in Cova (2022), there are cases in which the Commission does not seem to target just countries with objectively poor labour market and/or macroeconomic performance, as hypothesized in H_1a, but countries that are endowed with certain labour market institutions, such as having a minimum wage or highly centralized wage bargaining, which may appear to the Commission as hindering a country’s competitiveness. In these cases, the Commission seems to be less guided by performance indicators (as its role of ‘EU guardian’ would require) and more by general distrust towards those institutions standing in the way of labour market deregulation. In other words, the Commission is more ‘political’ in its scrutiny, thus displaying a prejudice against a number of pro-labour practices. For this reason, we hypothesize that:

H_1b: The Commission issues recommendations on wage policy based on institutional characteristics of national labour markets and wage-setting mechanisms.

As concerns the second stage of the policy process (the adoption of reforms), we hypothesize that distinct layers of supranational pressures have separate effects on countries (Guardiancich and Guidi, 2022). Simplifying, we distinguish between political pressures and financial pressures.

Supranational political pressures are represented by either the CSRs issued by the Commission, whose formulation we endeavour to explain in the first two hypotheses, or the existence of EAPs and SBAs signed with the troika and the IMF. The embodied conditionality varies and, hence, the two instruments need to be treated separately (see Mariotto, 2022; Efstathiou and Wolff, 2023).

As regards CSRs, which are minimally to moderately binding, we expect that reforms will follow the policy direction of recommendations—if the EU recommends an anti-labour policy change, anti-labour policy reform will be adopted, and vice versa.

H_2a: Wage policy reforms follow the direction of country-specific recommendations on wage policy.

Concerning EAPs and SBAs, which are agreed when a country is under severe pressure to adopt structural reforms (which usually imply the reduction of workers’ protection, see Supplementary Appendix Table A.2), we hypothesize that their presence will trigger anti-labour reforms:

H_2b: The presence of EAPs and SBAs triggers anti-labour wage policy reforms.

Supranational financial pressures are, essentially, the bond yields, which signal that financial markets are concerned with a country’s default risk and that, hence, they expect structural reforms to happen (for a discussion, see Efstathiou and Wolff, 2023). Following Guardiancich and Guidi (2022), we assume that domestic policymakers, faced with the heightened scrutiny of international finance, tend to decrease the protection accorded to their labour force. Hence, we hypothesize that:

H_2c: The higher a country’s sovereign bond yields, the higher the chance of anti-labour wage policy reforms.

5. Data and operationalization

Our dataset is built as follows (see Supplementary Appendix Table A.6). We start from the coding of country-specific recommendations and labour market reforms (the dependent variables for H₁ and H₂ respectively) as illustrated in Section 3. This allows us to construct a dataset with information on 28 countries (the actual EU-27 plus Great Britain, a member state until 2020) for 12 years, from 2008 to 2019, regarding reform events—336 observations in total. As far as CSRs are concerned, obviously, the dataset starts from the first year of the Semester (2011). Each country j in year t has a value based on the coding of CSRs calculated as follows:

where $s_{\mathit{ijt}}$ is the direction of a certain policy recommendation. More specifically, $s_{\mathit{ijt}} \in \left\{-1, 0, 1\right\}$⁠, and $s_{\mathit{ijt}}=-1$ if the policy prescription is anti-labour, $s_{\mathit{ijt}}=0$ if it is neutral, and $s_{\mathit{ijt}}=1$ if the prescription is pro-labour. N is the total number of recommendations received by country j in year t. For countries that do not receive any recommendations in a given year, $CS{R}_{jt}=0$⁠. In those cases when CSRs include more than one policy prescription (e.g. in 2011, Cyprus was asked to reform both the wage bargaining and wage indexation systems), these represent separate entries for our coding.

Similarly, for each country we have a value summarizing the strength and direction of the reform effort, calculated as:

where $c_{\mathit{wjt}}$ is a given component of a reform enacted by country j in year t. In the same fashion, $c_{\mathit{wjt}} \in \left\{-1, 0, 1\right\}$⁠, and $c_{\mathit{wjt}}=-1$ if the component decreases workers’ protection the labour market, $c_{\mathit{wjt}}=0$ if the component’s impact is neutral, and $c_{\mathit{wjt}}=1$ if the component increases workers’ protection. W is the total number of components of all reforms enacted by country j in year t. For countries that do not adopt any reform in a given year, $R_{jt}=0$⁠. The distribution across countries and years of the two dependent variables can be seen in Figure 1.

As for the operationalization of the explanatory variables, we list them in the order of the hypotheses. The ‘competitiveness indicators, labour market statistics and economic fundamentals’ of H_1a are the following:

• the percentage change compared to same period in the previous year of the Nominal Unit Labour Cost (NULC) based on persons (Eurostat, 2023a);

• the public wage bill, operationalized as the variable Compensation of employees (D1p) (UWCG) as percentage of GDP at current prices (Ecofin, 2023);

• the current account balance, expressed as a percentage of GDP (Eurostat, 2023b);

• the employment rate, as a percentage of the workforce aged 15–64 (Eurostat, 2023c).

The ‘institutional characteristics of national wage setting mechanisms’ of H_1b are the following:

• the minimum wage, operationalized as a dummy variable indicating whether the country has or does not have a minimum wage (Eurostat, 2023d);

• the centralization of collective bargaining, operationalized as the variable Central in the OECD/AIAS ICTWSS database (OECD and AIAS-HSI, 2023);

• the coordination of wage-setting, operationalized as the variable Coord in the OECD/AIAS ICTWSS database (OECD and AIAS-HSI, 2023).

One caveat is due. Ideally, the model should test for public sector NULC. Regrettably, Eurostat’s dataset ‘Labour cost levels by NACE Rev. 2 activity’ excludes section ‘O: Public administration and defence; compulsory social security’, most probably due to confidentiality issues.

Turning to the hypotheses aimed at explaining reforms, we operationalize ‘recommendations on wage policy’ (H_2a) with the variable CSR as described above (note that, when testing these hypotheses, the dependent variable of the first analysis becomes one of the explanatory variables), and the “presence of EAPs and SBAs” (H_2b) with a dummy variable scoring 1 if the country is under a programme supervised by supranational institutions in a given year and 0 otherwise (coded by the authors). To operationalize the ‘sovereign bond yield’ of H_2c, we employ the average yield of a country’s 10-year bond in a given year (Eurostat, 2023e), except for Estonia, whose values have been imputed using the package Amelia (Honaker et al., 2011) in R.

Other control variables that we use in the statistical analysis of are the following:

• GDP growth, as a percentage of GDP (Eurostat, 2023f);

• the country’s government ideology, calculated with data provided by ParlGov (Döring et al., 2023) as the mean of the left-right position of the parties that support a certain government (ranging from 0 at the extreme left to 10 at the extreme right), weighted by their number of seats;

• the country’s government position towards the EU, calculated in the same way as the variable above using the eu_anti_pro variable in ParlGov;

• a categorical variable indicating the different European Commission in office in the period of interest (Barroso I, Barroso II, Juncker).

6. Empirical analysis and discussion of results

Since we have time-series cross-sectional data, we test the hypotheses listed above with linear mixed-effects models in which we specify a list of covariates common to all observations (the variables described in the previous section) and random intercepts for countries and years. Formally, our models can be written as follows:

where $y_{jt}$ is the value of the dependent variable for country j in year t, ${\beta }_0$ is the general intercept (for all countries and years), the $\beta$’s are the coefficients of the country-year predictors x’s (with K predictors in total), $u_j$ is a set of 28 country-level intercepts, $v_t$ is a set of 12 year-level intercepts and $e_{jt}$ is the residual error term. $u_j$⁠, $v_t$ and $e_{jt}$ are assumed to be normally distributed with a mean of 0 and a constant variance. All regression models are fitted using a restricted maximum likelihood (REML) criterion with the function lmer in package lme4 (Bates et al., 2015) in R.

We begin with the results of the first set of models (illustrated in Figure 2), which are aimed at testing H_1a and H_1b. We run three nested models: in the first model (the blue one in Figure 2) only the variables aimed at testing H_1a and H_1b are included; in the second model (the red one) we add economic controls (GDP growth, budget balance and bond yields); in the third model (the green one) we further add political controls (Commission, government ideology and government stance towards the EU). Overall, the controls do not seem to add significant explanatory power to the model, as indicated by the higher Bayesian and Akaike Information Criterion of Models 2 and 3 compared to Model 1.

Out of the seven covariates included in Model 1, five are significantly correlated with the direction and strength of recommendations on wage settings. These are compensation of public employees, current account balance and employment rate (employed to test H_1a) on the one hand, and minimum wage and centralization of wage bargaining (employed to test H_1b) on the other. It is worth stressing that some indicators that we used as proxies for both hypotheses are not statistically significant in our models: the NULC, for instance, does not seem to be consistently taken into account by the Commission in CSRs; the same happens for coordination of wage bargaining (possibly because the Commission focuses on the easier-to-observe centralization only, see Note 5). Conversely, those variables whose coefficients are significant in all models are the ones for which we have the most robust evidence of a relationship with CSRs. To summarize the findings of the first set of models, we can state that the Commission tends to recommend restrictive wage policy measures to countries with a higher public wage bill, higher current account deficit and lower employment rate, as far as objective indicators are concerned, and to countries with a minimum wage and centralized collective bargaining, which pertain to the subset of ‘political’ indicators. If we want to be more conservative and trust only the covariates that we find as statistically significant in more than one model, public wage bill, current account balance and centralization of collective bargaining are the indicators that are most certainly related to the issuing of anti-labour CSRs in wage policy. Given that the public wage bill is preeminently treated as a fiscal rather than as a labour market indicator, the Commission seems to be more preoccupied with certain institutional characteristics of national wage-setting mechanisms, in line with Cova (2022), as well as with the cost of public employees than with the performance of the labour market stricto sensu.

In the second set of models, illustrated in Figure 3, we follow a similar approach as with the previous set, assessing first a restricted model with only the indicators aimed at testing H_2a, H_2b and H_2c, and then adding economic controls in the second model and political/institutional controls in the third model. Since the two factors that we expect to play a role in H_2a and H_2b (CSRs and EAPs) exclude each other—countries under an EAP do not receive CSRs—we make them interact with each other, so that the coefficient of $\mathit{CSR}$ estimates the effect of recommendations when $\mathit{EAP}=0$⁠, and the coefficient of $\mathit{EAP}$ estimates its effect when $\mathit{CSR}=0$⁠.

The positive coefficient of $\mathit{CSR}$ (which is significant in all the models, leading us to reject the null hypothesis for H_2a) indicates that, overall, pro-labour recommendations are associated with pro-labour reforms, and vice versa: even if they are not the only factor at play, CSRs influence reforms. Given that the Commission’s recommendations predominantly advocate for a reduction in workers’ protection, CSRs often justify an anti-labour stance in domestic wage policy. EAPs, instead, do not seem to have a significant effect, possibly because countries under an EAP experience, at the same time, high risk premia on their government bonds (the mean value of bond yields for countries under an EAP is 8.60, and 2.97 for the rest), the effect of which is strongly significant and its direction is unambiguously anti-labour. Instead, if the variable bond yields is removed from the model, the coefficient of EAP becomes statistically significant at <1 permille. Therefore, we reject the null hypothesis for H_2c but not for H_2b. Disentangling the interplay between EAPs and bond yields is, methodologically, a challenge. Since excluding one of the two would be problematic to justify, we simply report their estimates when both are included in the regression, interpreting them as an indication that financial pressures supersede political ones for countries under EAPs.

Among the economic controls, we find that a consistent and significant effect is exerted by GDP growth: countries tend to adopt expansionary reforms when their growth is higher. This suggests that restrictive wage-policy reforms have been, among other things, one of the ‘austerity tools’ employed by countries dealing with stagnant or negative economic growth. Among the political/institutional controls, a relevant role seems to be played by coordination in wage bargaining, which we, in line with several industrial relations scholars (see Kenworthy, 2001), interpret as a proxy of institutional capacity: higher coordination in wage bargaining increases the ability to impose austerity. Finally, the Commission variable tells us that the reform effort leading to restrictive wage policy has been stronger under the Barroso II Commission and weakest under the Barroso I Commission, with the Juncker Commission lying somewhere in between, in line with other readings of the developments in the EMU (see Braun et al., 2022).

We ran a number of alternative regression specifications as robustness checks. First, to better appreciate the change in the impact of the Semester over time, we have run the models shown in Figure 3 on two separate subsets of the dataset, one comprising the years between 2011 and 2014 (Supplementary Appendix Table A.9) and another including the years between 2015 and 2019 (Supplementary Appendix Table A.10). The results indicate that in the ‘first period’ reforms (which have been mostly anti-labour) have been pushed by CSRs and bond yields, while in the ‘second period’ reforms (mostly pro-labour) have taken place in countries with higher growth and lower public wage bills. We also tested the impact on reforms of the number and direction of recommendations included in MoUs. The results (Supplementary Appendix Table A.11) do not reveal any significant association, while the rest of the covariates, most notably CSRs and bond yields, remain statistically significant. Finally, we checked whether higher bond yields strengthen the impact of CSRs on reforms by means of an interaction between the two variables (Supplementary Appendix Table A.12). Being the interaction not significant, we exclude the existence of a reinforcing effect exerted by bond yields.

Overall, the statistical analysis presented in this section lends credence to a heightened supranational influence on national wage-policy reforms, chiefly leading to reduced workers’ protection. These pressures have been both political and financial in ‘ordinary times’, and mainly financial during ‘hard times’, when countries under fiscal strain had to comply with the troika or IMF recommendations.

7. Conclusions

Wage setting in European member states is a policy field where national sovereignty has been eroded in the aftermath of the global financial crisis, eliciting agreement among scholars on the institutional underpinnings and direction of such developments. Building on these foundations, this article innovates the field in three ways. Theoretically, it gathers evidence for an emerging multi-level policy cycle, showing how supranational political and financial pressures influence national-level decision-making. Empirically, the EU’s impact on domestic wage policy from the global financial crisis until 2019 is mapped through an integrated dataset of uniformly coded recommendations and reform events. Methodologically, the article makes it clear that, if researchers want to measure the impact of CSRs on reforms, they should focus on policy outputs instead of relying on the Commission’s evaluations as a proxy.

At the formulation stage of country-specific recommendations, our research refines and clarifies earlier work. The European Commission bases its advice on a mixture of institutional factors—the existence of minimum wages and, in line with Cova (2022), greater centralization of collective bargaining elicit more pressing demands for wage moderation—and public finance considerations—an expensive public wage bill requires greater rigour (see Di Carlo, 2023). Additionally, we cannot rule out that other macroeconomic indicators, such as the employment rate and the current account balance of a country, make the Commission more likely to demand reforms less favourable to labour. Hence, the findings of several qualitative studies are confirmed (see, among others, Armingeon and Baccaro, 2012; Scholz-Alvarado, 2022; Syrovatka, 2022): the Commission, at least during the sovereign debt crisis, saw the liberalization of the labour market as a tool to mitigate a poor performance in these indicators.

Irrespective of the underpinnings of CSRs, a more pressing question is to fill in the gap in the literature on the factors underscoring domestic reforms, a key condition to assess the characteristics of the nascent multi-level policy cycle. The analysis confirms a number of qualitative conjectures, mainly developed by Glassner and Keune (2010). Regressions unambiguously show that financial pressures conveyed through soaring bond yields played an autonomous role, transcending the political conditionality embodied in Economic Adjustment Programmes. Moreover, we find that the issuance of CSRs indeed triggers wage-setting reforms. In other words, the Commission influences (in part, at least) national policy in the desired direction. These effects seem to be dependent on the general economic situation: sustained GDP growth as well as current account surpluses steer countries away from austerity measures. Finally, institutional capacity matters. We find that, in contrast to the thesis that decentralized industrial relations are more conducive to reforms, greater coordination of labour market actors increases the ability to impose unpopular measures.

The impact of such supranationalization of wage policy on domestic decision-making should be a cause for alarm. Both European political and international financial institutions favour reductions over increases in the protection of workers. Their preferences then affect domestic policy reforms, leading to a weakening of labour market institutions and a reduction in the well-being of workers. In light of a recent shift in European priorities away from austerity, time will tell whether such anti-labour attitudes are a historical accident, circumscribed to the EMU’s first two decades in existence, or not.

The article opens up numerous avenues for further research. The most natural follow-up to this analysis would be exploring how the cycle ends, that is, whether the Commission’s evaluations of the implementation of its recommendations, contained in the yearly Country Reports, take into account these countries’ reform effort, and whether evaluations are more generous when reforms go in a certain direction. That being said, the breadth and richness of the CSRs and LabRef databases allows scholars to study related policy domains: the application of similar research methods to fields such as unemployment benefits, active labour market policies, labour taxation and so on, may constitute the basis for an overall assessment of the impact of the supranational pressures on national labour and social policies and beyond.

As our analysis focuses on the years preceding the COVID-19 pandemic, it would also be interesting to investigate whether that crisis has changed the attitudes of the EU and the member states. Given the focus on stimulating investment epitomized by the NextGenerationEU (NGEU) recovery package and the Commission’s recent pro-labour initiatives, the tendency to see internal devaluation as a growth-boosting strategy has most probably been dampened. Yet, as NGEU’s Recovery and Resilience Facility is embedded in the European Semester, whose conditionality has been strengthened, this may further tighten the European ‘grip’ on national policymaking in the not inconceivable possibility that anti-labour strategies are rekindled.

Acknowledgements

The authors would like to express their gratitude to Paul Copeland, Joshua Cova, Donato Di Carlo, Giovanni Esposito, Martin Fischer, Paolo Graziano, Matteo Jessoula, Manos Matsaganis, Guglielmo Meardi, Marcello Natili, Igor Tkalec, Višnja Vukov, Christian Welz, Jonathan Zeitlin and Tiziano Zgaga for their intellectual input and comments on earlier drafts as well as to Eugenio Borgognoni for his research assistance. Moreover, we are grateful to the participants to the various panels and workshops where the research was presented, that is, at the 36^th conference of the Società Italiana di Scienza Politica (SISP), the Political Economy Working Group of the European University Institute, the Centre d'Etude des Politiques et de l'Administration publique (CEPAP), Université libre de Bruxelles, the Welfare Seminar Series, Faculty of Political, Economic and Social Sciences, Università degli Studi di Milano, the Convegno Società Italiana di Sociologia Economica (SISEC) 2024, the Scuola Normale Superiore and the ViPET (Vienna Seminars in Political and Economic Transformations) at the University of Vienna.

Supplementary material

Supplementary material is available at Socio-Economic Review Journal online.

Funding

Part of this study has been funded through the project SocialESti—Socializing the European Semester through Formulation, Implementation and Evaluation, financed through the STARS—Supporting Talent in ReSearch programme of the University of Padova.

References