“To what extent is the European Council
responsive to the salience of unemployment?”
Capstone: Public Opinion and European Integration
Student: Giovanni Capitta
S2368978
Supervisor: Dimiter Toshkov Track: Economics and Governance
Master’s Thesis in Public Administration
Abstract:
Does public opinion have a causal effect on the European Council agenda in matters of employment policy? Extending on previous literature, I attempt to establish whether there exists a contemporaneous or lagged correlation between public opinion and the European Council agenda by running dynamic time-series analyses using different model specifications. Contrary to expectations, I find that the European Council agenda is unrelated both to public opinion and the EU level unemployment rate. On the other hand, the unemployment rate is a highly significant predictor of Eurobarometer prioritisation of unemployment, strengthening the claim that voters are rational along pocket-book lines. I also find that the ideological composition of the European Council exerts no independent effect on the representation of job concerns in the agenda. However, I find that the weight of the unemployment issue in election manifestos has increased consistently over the period of analysis (2000-2018). Lastly, I find no evidence of an informal agenda-setting role of the European Council president on employment policy over the period of analysis.
Acknowledgments
I would first like to thank my supervisor Dimiter Toshkov whose expertise was invaluable in guiding me through the writing process.
I would also like to thank my friends at Leiden – Jasper, Sidi and Mark – for the many opportunities of constructive debate they made possible during my first year in the Netherlands, particularly on my thesis topic. Most of all, I would like to thank Roniette for being by my side every step of the process despite her being in a different country.
Most of all, I want to thank my family for the constant support and patience they displayed during this year. In particular, I want to thank my mother and father for always being there when I needed them and my sister for her support especially in the last days of this project.
Introduction
...
4
Literature Review
...
7
Elections as a sanctioning mechanism
...
8
Imperfect Rationality of Voters
...
9
Empirical Evidence of the Policy-opinion Nexus
...
11
Political responsiveness in the EU
...
14
Dynamic responsiveness: a focus on agenda setting
...
15
Relationship between the agenda and policy outputs
...
18
The Case of the European Council
...
19
The European Council: a historical perspective
...
21
The Democratic deficit: long story short
...
23
Methodology
...
34
Dependent Variable
...
35
Measuring the ideological “Centre of Gravity”
...
38
The informal agenda-setting competencies of presidents 40
...
Model Estimation
...
44
Results and Analysis
...
46
Dynamic responsiveness: A visual analysis
...
46
Estimations: is there a relationship?
...
48
Error-correction models and an extension
...
51
Conclusion
...
55
References
...
60
Introduction
The global economic fracture triggered by the financial crisis, in the years following the sub-prime bust, has caused Europe to lose a decade of strong and consistent economic integration and development, whilst forcing a sundering rift among countries in income inequality and GDP growth. From these premises follows the 2020 strategic paper presented by the European Commission in 2014 aimed specifically at maintaining peace and cohesion, rebolstering economic growth, and solving the union’s fragilities by way of greater investments in technology and infrastructure. Among all objectives was to increase the employment rate among the 20-65 age bracket to 75% with a focus on women and migrant workers. To this day, this goal seems to be obtained at least for what concerns the unemployment rate (6.5% average equal to pre-crisis levels). However, vast differences across member states still exist: the lowest unemployment rate measured in the union is 2.8% in Czechia against 20% as measured in Greece (Lecerf and Dobreva, 2019) – signalling the missing coordination across member countries and the lack of a proper and systematic approach to positive integration in EU labour markets to meet the current challenges of globalisation.
The unemployment question in particular, has drawn many to believe that there is a disconnection between the verticalised structures of governance in the European Union and the public which is subjected to its rules: many define this as the “democratic deficit” of the European Union (Rovny, 2002), the result of which has been an increased presence of Eurosceptic parties particularly in those countries that have most suffered
no coincidence when the unemployment rate is demonstrably a strong predictor for Euroscepticism in recent literature (Sarassin et al., 2018). This is made even more evident by the ‘policy drift’ (Føllesdal and Hix, 2006) identified at the EU level, where the dynamics of legitimacy in EU decision-making seems not to take the demands of the public into enough consideration in the promotion of their agenda. However, the challenges tied to poorly legitimised governance structures die hard. Even after further democratisation of legislative and regulative processes, it cannot be assumed that legitimisation of governance at the EU level could be obtainable given the poor understanding that the electorate has of European institutional functions, which consequently implies that public preferences cannot act as a binding incentive for policymakers. It is in this context that the analysis of the relationship between EU policymaking and public demands, particularly in the realm of employment policy and job creation, becomes of vital interest to our understanding of contemporary democratic processes.
In this thesis, the focus will be placed on the agenda setting capabilities of the European council rather than the policy creation phase that follows any rhetorical deliberation. The thesis attempts to answer whether there exists a causal relationship from public opinion to the European Council agenda with the most adequate and scientifically verified datasets available today. The relationship between rhetorical responsiveness and public opinion has been investigated in different contexts (Bevan and Jennings, 2014; Alexandrova et al., 2016), with the latter one investigating whether the relationship between policy domain representations in the Council find significant correlations with public
opinion after having controlled for the so-called “intensity of the problem” (Alexandrova et al., 2016: 620). I am to take-off from their mixed findings and approach the question of agenda responsiveness focusing on the issue of job concerns and employment growth. Indeed, finding a relationship between agenda setting and public opinion on unemployment would raise the issue of why, if this relationship is robust, we do not observe more positive integration of labour markets and greater institutionalisation of the problem at the EU level. However, rejecting the hypothesis of a significant relationship would potentially raise an even greater concern: could it be that EU institutions are unresponsive to the demands of the public?
The thesis will be organised as follows: the first section will be aimed at summarising the recent literature on the topic in order to highlight the most significant debates concerning democratic responsiveness; the second section will outline the case of the European Union and the that of the European Council as well as putting forward the main hypotheses; the fourth section will be focused on the methodology and the data selection for the empirical analysis; the fifth section will comprise the results and an analysis as well as an extension or robustness check for the empirical analysis; a conclusion to the argument will then follow to summarise the findings and to highlight areas of interest and concern. Indeed, this thesis provides evidence of a lack of relationship, both contemporaneous and lagged, short-term and long-term, between public opinion and the European Council agenda. Quite surprisingly, I find that there is no relationship even for what concerns the intensity of the problem which would be usually assumed to be the signal marking
how strongly one issue should be prioritised over another. This finding could potentially signal a lack of democratic responsiveness in one of the most important institutions of the European Union, as well as raising questions about what, if not the intensity of the problem and public opinion on the matter, drives the Council’s Agenda on employment policy.
Literature Review
Political responsiveness, in the words of Dahl, represents a key characteristic of democracy (Dahl, 1971:1-2) when framed as the “congruence of collective public attitudes towards political issues with the policy preferences (…) of elected representatives” (Hobolt and Klemmemsen, 2005: 380). Historically, however, several political theorists have argued whether a single mode of responsiveness is required for democratic legitimacy or whether there could be conflicting understandings of democratic representation. It is debated whether democracy should be considered as the rule of the people or the rule for the people, precisely because despite the two should – theoretically – be produced in unison, the presence of one may lead to the exclusion of the other (Sharpf, 1999). In the last decades, plenty of attention has been devoted to outlining the relationship between public opinion and policy, although the major strands of research resort to outlining this relationship for the American context (Page & Shapiro, 1983; Stimson et al., 1995; Hakhverdian, 2010). Indeed, this relationship is further complicated in the European Union given its verticalised institutional environment. However, some efforts were dedicated in this direction with mixed, and sometimes conflicting, results (Hobolt and Klemmemsen, 2005; Bolstad,
2015; Toshkov, 2011; Alexandrova et al., 2016). All in all, the literature points to a general correlation between public opinion and policy, highlighting the importance of issue salience and issue ownership to the question of responsiveness, although the EU context remains a ‘hard case’ for scholars to gain any further insight due to threats of reverse causality and external omitted confounders (ibid., 2016).
Elections as a sanctioning mechanism
Spatial models of elections tend to emphasise how politicians promote platforms that capture the medium voter in order to maximise their chances of electoral success (Downs 1957). In this framework, political responsiveness assumes some degree of voter rationality in processing diagnostic information on policy to be interpreted and used in their decision process to punish or reward the incumbent; sometimes retrospectively when it is cheaper than scrutinising candidates at the time of election (Downs, 1957; Key, 1966). When this assumption holds, it follows that politicians will choose their optimal strategy based on their degree of electoral slack, where elections are a means through which voters hold the elected officials accountable to their preferences. We understand accountability formally such that “one person A is accountable to another, B, if (…) there is an understanding that A is obliged to act in some way on behalf of B and (…) B is empowered by some formal institution to sanction or reward B for her activities in this capacity” (Fearon, 1999: 55). Governments are therefore accountable insofar as they are understood to be acting on behalf of the people and for the people and can be re-elected or sanctioned based on how closely they
achieve the objectives laid out in their electoral campaigns. Elections therefore have a selection purpose – where voters realign powerful institutions like parliaments or presidencies with representatives of political platforms that are deemed most suitable for the current state of affairs – and a sanctioning purpose, whereby incompetent or inadequate politicians that had been previously elected are not given a second opportunity. The two mechanisms are not disconnected though, as is shown in adverse-selection models where bad-type politicians pool with good-types in order to win office (Besley, 2006). Further, and independently of the competence level of the incumbent politician, electoral pressure induces both good and bad types to respond to voter preferences in fear of retrospective punishment at the polls (Fiorina, 1981; Barro, 1973; Ferejohn, 1986). If politicians are concerned with their re-election and are aware of how voters react to their policy proposals they will be “constantly piloted by the anticipation of those reactions” (Sartori, 1976:84), a mechanism that is also termed of “rational anticipation” (Stimson et al., 1995).
Imperfect Rationality of Voters
Taking the theoretical conclusions at face value, however, would induce one into believing that policymaking always converges to the most popular platforms, and that elite political decisions are always congruent with public opinion. Whether this is desirable or not is beyond the scope of this thesis. However, several scholars have brought forward two strong and opposite arguments to that of standard democratic responsiveness: (1)
bounded rationality of voters and (2) elite manipulation of information sources.
One of the arguments that is most prominent in the literature resorts to the idea that voters do not have stable political preferences and do not have the tools necessary to make their way through policy proposals to consistently understand information signals on policy, making it impossible for them to meaningfully respond to policy change. In many ways Converse (1964) is among the precursors of this idea by referring to belief-systems, and in particular the voter’s ordering of issues according to some degree of ideological congruence. He instead finds that voters tend to form opinions on certain specific issues, but then struggle to connect such issues together in a uniquely consistent view. On the other hand, some still maintain that individual voters may react to policy signals but cannot altogether distinguish which signals the politician should be held accountable for and which should be viewed as exogenous. In the words of Duch and Stevenson, “voters observe shocks to the macroeconomy but cannot observe the mix of exogenous and competence components that comprise these shocks” (Duch and Stevenson, 2008:30). If voters cannot separate competence signals from exogenous ones, then we would struggle to understand why Governments should concern themselves with responding to variation in public opinion when their punishment, or re-election, is determined by factors out of their control.
Conversely, even when voters can observe policy signals and construct a consistent opinion on the state of affairs, some argue that their opinion is still being manipulated by elites that control and dispose of the information they use. In this sense, voters are imperfectly rational because
the information they receive is not of sufficiently high quality to grasp the state of affairs, which then follows into them being unable to respond to government action. For instance, Iyengar (1991) shows evidence of how differences in the treatment of the poverty issue causes significant changes in the way public opinion attributes responsibility for poverty, and therefore its overall policy position. Some have also gone as far as arguing that the representation of public opinoin may have an effect on public opinion itself by producing bandwagon effects in election rounds (Ginsberg, 1986; Rothschild and Malhotra, 2014). Moreover, Page and Shapiro (1992) demonstrate that, despite public preferences remain coherent over time, at least for the U.S. context, there are instances where presidents can temporarily and significantly influence public opinion to their advantage. Therefore, the relationship or linkage between public opinion and public policy may be artificial when demonstrably piloted by third-party influences.
Empirical Evidence of the Policy-opinion Nexus
Several articles have attempted to outline the empirical relationship between public policy and public opinion, successfully for the most part, although with some complications that have often led to measurements of weak correlations, providing little evidence in favour of effective responsiveness between public opinion and public policy.
The methods through which this relationship is investigated use two main types of research designs: static and dynamic. An example of a static design could be shown in Miller and Stokes (1963), where the authors use cumulative scaling on fixed policy dimensions and observe
the correlations between constituency preferences and representative votes. They find substantial correlations, particularly on matters such as “social welfare” and “rights of black people”, although most of them are not statistically significant. As opposed to micro-level studies, research like that of Weissberg (1978) attempts to move from the “dyadic” into the “collective” to account for more “refined degrees of representation” (Weissberg, 1978: 540) by focusing on actual policy rather than mere vote intentions. In a similar vein, Erikson (1978) attempts to re-examine the weak relationship found by Miller and Stokes (1963), finding that their results seemingly underestimate the correlations between constituency opinions and congressional behaviour. Another important study along the same lines but focusing more on macro-level cross-state variation was that of Erikson et al. (1993), which investigate cross-sectional relationships between public opinion and policy-making using disaggregated poll data, finding significant evidence for state-specific policy positions feeding into policy outputs. However, static designs as the ones outlined above will all struggle to move beyond simple correlations due to the common threat of reverse causality, meaning their value is more anecdotal than explanatory of a presumed causal relationship. In addition, studies of this sort tend to lack appropriate controls for mediating or confounding factors that could either be related to one, the other or both of the variables of interest.
Another strand of literature has dealt with dynamic responsiveness using more modern, multivariate, time-series approaches. Above all, the most influential and pioneering study is the one by Page and Shapiro (1983) that investigates the covariational congruence of public opinion and
policy over 357 instances in the U.S. using one-year time-lags. After having analysed all their no-change scenarios they conclude that policy changes are congruent with opinion changes over 66% of the time. However, they also acknowledge that many cases cannot be taken to simply fit into one category due to ceiling and floor effects that prohibit policy to adjust to popular preferences (Page and Shapiro, 1983: 178). Another important article in a similar direction is that by Stimson et al. (1995), which estimates a Kalman filter linear model to match “public mood” and “policy activity” over time, finding a 0.36 percentage point change in the overall policy measure for a one percentage-point change in public opinion. Furthermore, Wlezien and Soroka (2012) provide an empirical extension of their initial “thermostat model” of democratic representation that was introduced at an earlier stage (see for example Wlezien 1995), with a specific emphasis on survey data and in particular the ISSP to analyse the relationship in a broader set of OECD countries. The famous model refers to the idea that relative policy preferences are a function of the difference between preferred and current policies; governments will tend to respond more intensely to public opinion the greater the difference in the parameters observed over the entire electoral cycle (also see Soroka and Wlezien, 2010). In a similar style, also Hill & Hurley (1999) outline how there is some degree of reciprocity in the relationship between public opinion and policy when considering spending measures.
Political responsiveness in the EU
Some articles have also targeted the EU context more specifically, by analysing the effect of public opinion on both policy measures and party positions. Of the many, Spoon and Klüver (2014) investigate the issue responsiveness of political parties to changes in public opinion, finding that parties increase attention to an issue in the order of 0.08 percentage points following a 1 percentage point increase in public attention. Further, they find that this issue responsiveness is even more significant when European elections are in close proximity to national elections (Spoon and Klüver, 2014: 55). In a similar vein, Spoon and Williams (2017) investigate the incidence of eurosceptic public opinion on the salience of eurosceptic positions in EU parliament election manifestos. They find that as the former increases in intensity the presence of the latter increases in EU election manifestos, and that this relationship is likely conditioned by the size of the party in question. A strand of the literature has focused on specific policy outputs from particular EU institutions such as the European Council, the Council of European Union or the European Commission. In particular, Wratil (2015) investigates the degree of democratic responsiveness in the Council of the European Union by resorting to the “decision making in the European Union dataset” (p.14), finding that the willingness of governments to present positions congruent with public opinion will vary with the electoral pressure they are subjected to. In fact, governments are indeed more responsive to left-right public opinion as national elections near, as well as under majoritarian electoral rules, although the coefficient is only weakly significant (p.23).
Some have also focused more specifically on the EU commission, albeit in a very different fashion. Rauh (2019) investigates how politicisation of particular issues renders the European Commission more sensitive to them when these receive public attention. He observes 17 policy draft proposals between 1999 and 2008 and finds that the European Commission actually does respond to high salience with increased consumer protection. On the other hand, Haverland et al. (2016) conduct the first and broadest investigation of special Eurobarometer surveys, finding that the European Commission tends not to ask questions on matters where it has extensive competencies, such as “external trade” and “competition”. Indeed, they also find that there has never been a special Eurobarometer survey on the issue of immigration (at least up to the date of the article), hinting at how the European commission may be capable of forming public opinion itself through the selection of topics asked in its surveys.
Dynamic responsiveness: a focus on agenda setting
While the possibility of a direct transmission of public opinion into policy outputs is unlikely, it still might be the case that priorities in government be aligned with public opinion for what concerns the macro-topics that have to be engaged with in policy-making. It is plausible that elite politicians will still debate issues that the public considers of overriding importance without necessarily acting upon such issues in the way that the public expects. However, it may also be that the organisation in charge sets its priorities according to different criteria, where the public’s formal prioritisation is but one of the available signals. The connection between
organisational priorities and outputs is a very well-researched and established field of study, yet the internal dynamics through which these priorities become policy outputs, and more so the relationship between public salience and agenda-setting, is an area of this literature that has yet to be understood in its entirety.
According to most readings, agenda setting refers to the process through which organisations shift attention from one issue to another in a context where attention is a scarce resource (Jones and Baumgartner, 2005: 39). In his pioneering work, Herbert Simon (1983) defined two critical ways in which individuals, as well as organisations can process information: serial processing – which refers to dealing with individual problems one at a time, shifting to a separate task only once the previous one is dealt with, and parallel processing abilities – which instead refers to what is more commonly known as multi-tasking, the ability to deal with more problems at once. Organisations have an advantage compared to individuals, in that they are capable of adapting to stimuli by setting up individual sub-systems that focus on a specific problem independently of the general organisation. Individuals struggle to do so with the same intensity, although multi-tasking in individual behaviour is still possible. Yet, and especially for unexpected changes in the external environment, organisations and human beings alike, will have to switch from parallel processing to processing in seriatim (ibid, 2005:39). This will be the case because despite organisations can be large and complex with several underlying facilities and layers of varied expertise, inadvertent changes from the external environment will usually have to be assessed by decision makers in the organisation subject to what Simon defined as a
“bottleneck” of attention, where only a finite amount of information can be processed at once (Jones, 2002: 276).
This particular trait of organisations has extensive consequences for what concerns policy change and, as will become clear, democratic responsiveness. The inherent bounded rationality of organisations reveals that these will struggle to react with policy in a way that is proportionate to the signal they receive. This will be the case when issues that the organisation deals with represent themselves consistently over time. Instead of increasing attention to the issue in a manner equal or proportionate to the signal they receive, organisations will usually tend to be biased towards previous modes of solving the problem when the problem definition is constant. Organisations will largely resort to similar solutions when faced with similar problem definitions, a characteristic that Pierson (2000) defines as “path dependence”. Governments and organisations, therefore, will under-respond to external information until particular thresholds are reached, when they will be forced to adapt to change by overreacting to the problem at hand. This characteristic makes organisations “disproportionate information-processors” (Jones and Baumgartner, 2005: 50), unresponsive to information signals until these cause a discontinuity in the problem definition of the issue itself. This characteristic of policy change then leads us to expect policy agendas to be largely constant and unchanged over time, as previous characterisations of policy problems or “status quo biases” render policy stable against external signals, and punctuated by stark and rapid increases in attention and policy diversion when the characterisation of a problem changes, or
when the severity of the issue at hand reaches a critical point – mindless of what this critical point may be (ibid, 2005: 53).
Relationship between the agenda and policy outputs
The relationship between the prominence of an issue on the agenda and its emphasis in policy outputs has been outlined quite extensively in studies of national budgeting (see Padgett, 1980; Carpenter, 1996; Baumgartner et al., 2006). However, the question of issue priorities in government aligning to those of the public is significantly more interesting to the democratic responsiveness problem. The reason for this is apparently straightforward: prominence on the agenda gives a positional emphasis to particular matters vis-à-vis others, increasing the probability of those issues represented on the agenda to be dealt with in the policymaking arena. Therefore, we should expect changes to policy in response to shifts in public opinion to be traceable to some degree by changes in emphasis in particular issues on the agenda of political organisations (Jones et al. 2009). It can be argued, therefore, that agenda representation, or otherwise agenda-responsiveness, of particular institutional arrangements will mediate the relationship between public opinion and policy outputs. Several scholars have attempted to outline this relationship rather successfully; Bevan and Jennings (2014) analysed the relationship between specific measures of UK and U.S. agendas – legislative, executive and budgetary – to changes in public opinion with a specific emphasis on the sources of friction in each of the different agendas. Indeed, they find that the lower the friction (which they define by appealing to institutional veto points) the greater the responsiveness between public opinion and policy outputs,
whilst the opposite occurs in higher friction environments. This finding is largely replicated by Bonafont and Palau (2011) in the context of Spanish politics, although they also find support for the decentralisation hypothesis (previously discussed by Soroka and Wlezien, 2010) such that the greater the degree of political decentralisation the stronger the correspondence of public policy and opinion. Further, Jennings and John (2009) also attempt to measure the congruence between the Queen’s speeches, as coded using the Policy Agenda project (Baumgartner, 2019), finding that the Queen’s speeches are affected by both short-run and long-run changes in public opinion for matters such as unemployment and the macroeconomy. Overall, there seems to be substantial evidence in favour of political responsiveness of political elites to changes in public opinion, although scholars tend to emphasise how institutional characteristics of those organisations subjected to changes in public opinion are relevant to how responsive these are on the whole.
The Case of the European Council
Indeed, the literature on democratic responsiveness, and particularly that which is at the intersection with agenda-setting theory, has focused greatly on governmental institutions at the national level more than at the international or supranational, and this is by no coincidence. In fact, as was previously mentioned, democratic responsiveness requires a sanctioning mechanism that is ensured by the political accountability of politicians to their electorates. This feature is arguably absent in many of the governing EU institutions that are therefore liable to misrepresent the will of their underlying electorate in the policymaking arenas. Indeed, this
thesis wants to extend on the work done by previous scholars on the question of political responsiveness in the EU and particularly with an emphasis on the European Council. Of the vast literature on the topic this particular question was only tackled by Alexandrova et al. (2016), that looked at the synchronic and diachronic associations between Eurobarometer public opinion on particular topics and the number of mentions on the European Council conclusions as a measure of the Council’s agenda. They find some evidence for democratic responsiveness but do find the effects to be very small, justifying this with the fact that the time-series observed only covered ten years and the non-stationarity of most of them made modelling with more complex specifications inadequate (Alexandrova et al., 2016: 617). Indeed, in this thesis we attempt to take-off from their findings and delve into the analysis of moderating variables that could modify the effect of public opinion or independently exert influence on the European Council agenda. This will be done by looking at an issue for which relatively longer time series data is available, namely that of unemployment. Indeed, as will be shown below, there are mainly two arguments that could be raised in objection of this thesis: (1) that the EU suffers a democratic deficit, and therefore that the underlying institutions below it will function such as to not allow proper accountability of policy-making to the electorate, and (2) that the issue of unemployment is largely a domestic issue, that is not dealt with at EU level, such that the observation of mentions of the issue at the EU level with increases in prioritisation of the issue should not be expected. I attempt to address each point in turn. Indeed, the following section will be organised as follows; the first section will provide a historical
characterisation of the European Council, its competencies and its development into one of the most influential institutions at the EU level. Then, I will respond to both of these objections and argue why the case at hand is one of interest and should yield the expected predictions. Lastly, I will explain two extensions of the argument in the form of moderating hypotheses to be tested contemporarily with those of pure democratic responsiveness.
The European Council: a historical perspective
The European Council is an organ of European governance that was first established in 1975 and was only formalised with the ratification of the Treaty of Lisbon in 2009. It is composed of the heads of state from each of the member states in the European Union, the President of the European Commission and the High Representative of the Union to Foreign Affairs and Security Policy, if and when matters of foreign policy are looked upon. When the European Council was set up, its formal competencies were not specifically drawn out but left open, with the sole purpose of encouraging consistent policy development among national states that had, until then, mainly pursued their own self-interest (Nugent, 2017: 183). As the tendency towards intergovernmentalism in European integration grew prior and following the Maastricht Treaty, the role of the European Council grew from its already substantial historical importance – see for example crucial summits like that of 1984 at Fontainebleau and its ramifications for the future role of the European Commission – to become a stage of alignment of national strategies, in the creation of common goals and priorities (Bickerton et al., 2015: 25). In time the European Council
developed into a very powerful agenda setter, despite it not having any kind of formal legislative power. The ordinary legislative procedure is what is known as the Co-decision procedure, whereby the European Commission presents a policy proposal or piece of legislation to the Council of the European Union and the European Parliament to consult on a par. Despite the significant increase in legislating power that both the Parliament and Council have garnered over the past decades, it is without doubt that the most important agenda-setting influence – at least from the Maastricht Treaty onwards – is the European Council. Despite its lack of de jure legislative power, it has de facto influence on the legislating facilities of both the Commission, the Council and Parliament due to the inevitably influential power relationship that the European Council enjoys with respect to the others. One instance in which the European Council was particularly influential was in the drafting of the 2014 Strategic Agenda from which, among other things, the 2016 tax-avoidance directives were introduced (Puetter, 2014). The European Council has overall grown its array of competencies, that now comprise high profile emergency issues like terrorism, but also expanded its influence to all macroeconomic debates as well as social policy, cohesion and enlargement (ibid, 2014:71). Despite not producing actual drafts, the European Council does in fact produce a document called the ‘Conclusions’, in which the heads of state set forward common priorities for both domestic and European policy-making – a document viewed by many as a valuable instrument to measure the EU agenda as a whole (Alexandrova et al., 2014). Indeed, considering the influence the European Council has on European policy, and the fact that all members of the Council are reflective of domestic
governments, it should follow that the European Council agenda be responsive to changes in the issue priorities of its underlying electorate. This should be even more true of the unemployment issue, being it a known fact that more salient issues lead to more responsive governments (Burstein, 2003: 30). However, as was previously outlined, there are two potential arguments that could make this hypothesis fail: (1) the fact that the EU suffers from a democratic deficit, and (2) that unemployment is, by and large, a responsibility of national governments. I will object to both of these arguments in turn.
The Democratic deficit: long story short
It’s been commonplace both in media and among scholarship to view the European Union represented as an illegitimate, supranational legislator. This is mainly because of the multi-level governance that controls European level policy entrusted to politicians that are very thinly accountable to the electorate. Indeed, there are several reasons why, among member states, there has been the widespread idea that the EU suffers from a democratic deficit. A concise definition of the term is hard to find, although most commentators refer to it by presenting some of the following arguments: (1) the European Parliament lacks sufficient legislative power, (2) there is a psychological barrier between Europeans and their institutions that creates distance between them, (3) that the European elections are mostly fought on domestic issues and are seen as second-order elections, and (4) that European integration produces policy drift from voters’ policy preferences (Føllesdal and Hix, p.534-536).
Some of those arguments have significant consequences on the theory underlying political responsiveness, in particular for what concerns the European Council. Indeed, the European Council is comprised of heads of state that are voted into power, mainly in light of domestic policy considerations. Some argue that the electorate is not aware of EU policy and that the EU Parliament elections are merely viewed as “second-order national elections” (Reif and Schmitt, 1980), seemingly because prior to the introduction of the Co-decision procedure – and even thereon – the results of the election would have virtually no impact on EU policy-making, rendering it a potential opportunity to punish national incumbents (Franklin et al., 1995). Further, even when policy change is introduced, voters are seemingly not responsive to changes in policy, since they cannot clearly discern or understand how and what policy is put forward at the European level. If voters are not responsive to policy, then it is unclear whether policy-makers would ever be concerned with bringing forth voter priorities rather than simply dealing with the problems they observe.
In the context of unemployment therefore, it would perhaps be more convincing that the European Council focused exclusively on the unemployment rate as objectively given rather than to what extent unemployment is viewed as a priority by European citizens. Furthermore, the idea grounding this research design is further complicated by a second point, namely, that the issue of unemployment is not strictly one that is dealt with at the EU level, since domestic labour markets are mainly regulated nationally. The main argument is concerned with the idea that most of the unemployment in member states is justified by other concerns, such as productivity, labour unit costs, job destruction rates and red-tape
regulation, which are likely to be dependent on domestic policy-making. As such, it is unlikely that the issue of unemployment be presented as a priority in the European Council mainly because its causes stem more from domestic factors than supranational intervention or regulation (Blanchard, 2005).
However, I think the two arguments are fundamentally problematic. Firstly, the idea that a democratic deficit exists as a result of there being a weak parliament begs the question whether elections are the only means of accountability for policy-makers, a claim in direct opposition to that which is theorised by most supporters of the “output-legitimisation” thesis (Majone, 1996; Moravcik, 2002). Furthermore, even if we concede that electoral accountability is a condition sine qua non for legitimate policy-making, the claim that voters choose their electoral strategy independently of the European state of affairs has become hard to believe. Several scholars have noted that voters, in fact, are indeed responsive to issues at the EU level and even consider European issues in their electoral choice at the poll station. Gabel (2000), for example, mentions the historical introduction of the Common Agricultural policy by Charles de Gaulle as a prime example of how national issues can be placed on the European agenda – in this particular case “(…) as a means to attract and maintain the farmer’s vote in France” (ibid, 2000: 52). Further, some have argued that issues such as EU enlargement – and in particular the accession of Turkey into the EU – were of fundamental importance during the German elections of 2005, such that voters were more likely to vote for candidates that held a position congruent with their own views on this issue (Schoen, 2008).
As a second point, one could argue that even if we granted the first point, and that is, that European voters have narrow considerations that do not take into account the EU’s broader interests, this would still not question the validity of an underlying hypothesis of agenda-responsiveness. This because even if voters are not responsive to the issue of unemployment at the EU level, they will surely respond to the ramifications of policy made at the EU level, or its absence, once it reaches the national arena. Government members occupy significant positions at the EU level, in the European Council but also in the Council of the European Union and Parliament. As a consequence of this, politicians that work at the EU level should be as accountable to the electorate for their performance in Brussels as for their performance at home (Moravcik, 2002: 612). As such, non-intervention in alleviating unemployment or a failure at doing so would still cause reactions from the electorate along pocket-book lines (Kinder and Kiewiet, 1981).
Turning to what concerns the unemployment issue, while there is some truth in the statement that national labour markets are left to be regulated domestically, it is equally true – and particularly for what concerns Eurozone countries – that many of the fundamental tools apt at regulating output and inflation, which are related to unemployment through Okun’s law and the Phillips curve, have been delegated to the European Central Bank in the adoption of a single-currency with fixed and irrevocable exchange rates which control is now under the politically insulated European Central Bank (De Grauwe and Ji, 2013). Equally, the Commission has tremendous influence on the unemployment problem and particularly youth unemployment and social policy through the
allocation of funds from the Multiannual Financial Framework. There has been some development in terms of unemployment-focused legislation, mainly for what concerns young workers in the ‘Youth Unemployment Initiative’, that was extended until 2019 (Cuperus et al., 2019) and also for what concerns increases in job and training expenditures in the EU budget. On top of this, discussions in the EU Council could produce a more intergovernmental approach, of coordination in labour regulation as well as of migration policy; another aspect which is seen as a vital shock-absorber in case of unemployment surges (Blanchard, 2005). Indeed, despite it not being strictly in its competencies, it is easy to see how many tools and opportunities the Eurozone and European Union combined may have in alleviating the problem, such that it becomes worthwhile to question why several initiatives have not even been placed on the Union’s agenda over the course of time.
Overall, there seems to be no doubt that the expectation concerning agenda-responsiveness of the European Council to the issue of unemployment is acceptable, both in light of the obvious accountability that the members of the Council have towards their national electorates, and also because the issue of unemployment is both one of high salience, enough to generate electoral costs domestically, and one which is to some degree under the control of the Union. Therefore:
H1: The greater the level of saliency of unemployment in public opinion the more representation job concerns will receive in the European Council conclusions.
A second issue that is worth emphasising concerns how the composition of the European Council could have an independent effect on the number of relative mentions which the unemployment question receives in the conclusions. This is because, despite the agenda of the Council being established in large part by its president, the corresponding conclusions and the emphasis that is placed upon the diverse issues presented in the summits will in large part depend on the resulting debate over problem definitions that occurs amongst policy-makers. Therefore, the composition of the Council’s constituent members is likely going to exert independent influence on the emphasis which is placed on this particular issue. However, the way in which this could impact the agenda-setting process of the institution writ large has to this day not been investigated, although some have made effort in this direction by considering both the electoral rule of constituent countries, and the electoral pressure induced by the proximity of elections (Wratil, 2015; Hobolt and Klemmensen, 2007). I wish to argue in a different direction, and that is that the parties represented in the Council are quite fundamentally linked to the agenda-setting process within it. I argue this to be the case by considering how parties voted into government are voted specifically because of the stances they advertise in their manifestos, and these become decisive at the poll station when they capture popular sentiment. It is hard to deny that political parties have historically meant more than the simple aggregate of their policy proposals (Bélanger and Meguid, 2008). This because parties develop and nurture underlying constituencies and socio-demographic clusters that are specific to their policy stances, to the point that they construct a self-image or “issue reputation” (Petrocik, 1996) that acts as a
signal for the voters and electorates that seek to represent those same policy stances. There are several instances of this concept in most OECD countries, most importantly in the ideological cleavage between the Democratic and Republican Party in the US, but also between the Conservative and Labour Party in the UK and the consequences this has on the positioning of particular issues on the legislative agenda (Petrocik, 1996; Green and Jennings, 2019). Indeed, we can understand this ideological division amongst parties not only through a historical and sociological lens, but also by accounting for the gains that particular parties obtain through specialising in the representation of particular interests. According to the Theory of Issue Ownership, parties will be expected to represent more vehemently those interests and issues that they deem to have a positive reputation on, in order to maintain credibility and reliability in the eyes of voters that have, over the course of time, developed a particular idea of the party’s image and characteristics (Petrocik, 1996). This is effectively driven by the idea that voters will wish to have specific issues represented by those parties that are most reliable in dealing with them, and this reputational trait leads “candidates [to] argue along lines that play to the issue strength of their party, and sidestep their opponents issue assets” (ibid, 1996: 828).
In this particular context, the issue of unemployment is one that has been historically the battlefield of several disputes along right-left cleavages (Hix, 1999) – particularly by proponents of left-wing, social-democratic policy-platforms and advocates of full-employment and Keynesian economics. Hence, we would imagine that a larger host of left-wing leaders occupying positions in the Council will increase the attention
and mentions of unemployment or job concerns overall. This expectation is reinforced by the fact that scholars have seen close relationships between the saliency of an issue and its ownership from parties (Walgrave and De Swert, 2007). In particular, Bélanger and Meguid (2008) go as far as to argue that issue saliency has a conditional effect on issue ownership, where parties will only aim to construct issue reputations upon issues that are salient to the public. I would expect unemployment to be one of those issues that politicians will be motivated to represent, specifically because it is easily comprehensible and visible across all segments of the electorate.
However, some could object that measuring unemployment concerns by simply relying on partisanal considerations (or a blind left-right comparisons) could be unjustified on two grounds: (1) there is some margin of error in probing whether parties can or cannot be defined as left-wing, since definitions have drifted away from ‘pure’ class conflicts to also include pro and anti-globalisation cleavages; and (2) even when it can be specified whether parties pertain to the left or right, it occurs that some of the heads of state are representing coalition governments that propose wide-ranging policy platforms. This makes it hard to make scientific claims on whether or not parties represent specific partisanal interests. It would appear that the definition of left-wing is either too narrow (focusing only on unemployment and power resources) or too broad (all interests that also include unemployment). On the other hand, it can still be argued that those governing parties that were elected with pro-employment platforms will attempt to place the issue of unpro-employment more forcefully on the agenda, mindless of their specific ideological affiliation. Several scholars have outlined how ownership is not entirely
fixed over time, particularly when certain issues become exceptionally prominent – such as during a financial crisis or humanitarian emergency – leading parties to reframe them in a manner that is convenient to the party or outright occupy them when they have not been fully absorbed by others (Holian, 2004; Blomquist and Green-Pedersen, 2004). It is not uncommon in fact, and particularly in recent years, to observe populist parties focusing on broad consensus topics, such as unemployment, specifically to gain political momentum from left and right wing constituencies alike. Therefore, we can expect that:
H2: The more unemployment-focused the ideological centre of gravity in the European Council, the more likely it will be for the European Council to mention the issue of unemployment in its conclusions.
It is plausible to assume that the nationality of the President of the European Council at different points in time will have varying effects on the amount of representation that the issue of unemployment receives on the EU Council agenda; namely, whether the country of origin of the president is at that time suffering above, below, or average unemployment. Following the Treaty of Lisbon, the President of the European Council was conceded a permanent role as opposed to the Council of Ministers. Some view this change to have strengthened the role of the president greatly in the face of the President of the European Commission, while others criticised this choice claiming it would reduce the ability for national governments to voice their concerns in the most apical agenda-setter for Europe and its institutions (Princen and Rhinard,
2006). Overall, the powerful position that the Presidency grants to particular national states is close to undeniable. In their formal roles, presidents can introduce political initiatives, and structure the agenda in such a way as to emphasise particular points instead of others (Alexandrova and Timmermans, 2013). Similarly, presidents act as moderators during the Council discussions, choosing how much attention must be dedicated to different issues, which then means more or less emphasis will be put on those issues in the Council conclusions (Ferreira-Pereira, 2008). This may lead one to expect Council presidents to exploit this arena and their role to further domestic rather than EU interests.
In a similar way, the presidency of the Council of Ministers can also be expected to have an indirect influence on the matters that are discussed in the European Council. In large part, this influence stems from the institutional configuration existing prior to the introduction of the Lisbon Treaty, where presidencies were synchronised, and the Council of Ministers acted as an operational arm of the European Council. Following the ratification of the Treaty in 2009, the Council of Ministers maintained this set up with a six-month rotation period for each country in the Union. Given the very short time that countries have in charge of the Presidency, and conscious that each country will more or less have one Presidency per decade, it is imaginable that there will still be significant spill-overs in priorities from the Council of Ministers into the European Council for two main reasons: (1) because the ministers represented in the Council holding presidency have power to set the agenda in the different configurations of the Council, which will lead to differing problem definitions and, consequently, differing priorities put forward in the European Council,
and (2) because under the co-decision procedure it is only the Council of Ministers that can put forward amendments, so many points of agreement reached in the Council of Ministers can be picked up by the European Council on a rhetorical level.
On the contrary, some could argue that the European Council should dismiss domestic interests in favour of the collective European good. In fact, it is in the very handbook of the Presidency of the European Council that the role is outlined to be “neutral and impartial” meaning the President “(…) cannot favour its own preferences or those of its member states” (General Secretariat of the Council, 2018). In fact, several scholars have outlined how a specific concern on national, or domestic, interests could rule to be counterproductive in the EU arena. In particular, Princen (2009) outlines how effective agenda setting at the EU level requires, among other things, the framing of domestic issues as a European-level issue, rather than passing as a merely national interest.
However, this contention is not as extreme or damaging to the general hypothesis as one might think. First and foremost, the presidential handbook provides a general guideline of how the President should act rather than a strictly prescriptive list of commands. In fact, there is no sanction or punishment against acting otherwise, although a behaviour that is too self-interested could prove counterproductive towards “consensus building and effectiveness in joint-action” (Alexandrova and Timmermans, 2013: 321). And even then, some issues would struggle to be framed specifically as national interests, since they are common to many countries in the Union to varying degrees. Indeed, the issue of unemployment would appear to be one such issue. Therefore, it does
appear sensible to assume that presidents that are more committed towards presenting the unemployment issue at the EU level should have the power to do so through the channels aforementioned. But which presidents are more likely to be committed to the unemployment problem? Perhaps the ones that receive more pressure domestically to raise such a problem at the European level; namely, those that suffer above average unemployment at home. The third and last hypothesis would therefore be that:
H3: The greater (more positive) the difference between the recorded unemployment rate in the country holding the Council presidency and the EU-28 average, the more responsive the agenda will be to the issue of unemployment.
Methodology
In order to empirically assess the validity of the relationships mentioned above, I will resort to investigating the causal impact of public opinion on the rhetorical positions of European Council members over time – that is, on rhetorical responsiveness in the European Council to changes in public opinion (Hobolt and Klemmensen, 2007). The methodological approach takes off from the most recent and extensive research project on European Council agenda responsiveness by Alexandrova et al. (2016) and attempts to extend the analysis to other predictors in order to deepen the understanding of the agenda-setting process.
Dependent Variable
As a measurement of the European Council agenda, I resort to the Comparative Agenda Project (CAP): a relatively new database that has coded the official conclusions from the European Council meeting consistently from 1975 to 2014. As previously mentioned, constructing a measurement of agenda largely entails capturing the attention or serious consideration of policy-makers for particular issues (Kingdon, 2003:3); and in this case, to specific issues that are discussed in the European Council setting over time. In this respect, the conclusions present a very useful barometer of the agenda, since they provide a comprehensive overview of the major points and priorities for the European Council and its constituent leaders for the upcoming months. Within the conclusions separate issues are allocated a particular code among a choice of almost 200 available (Alexandrova et al., 2013) measured at the quasi-sentence level, providing the most granular source of agenda measurement available to date. This allows precise monitoring of the issues mentioned in dynamic perspective enabling the “retrospective reconstruction of agendas” (ibid, 2013: 155). However, there are two main drawbacks from using this kind of dataset: (1) that the measurements of issue salience are not directional, meaning they are not measured in intensity, and (2) that the conclusions may contain “a mixture of substantive and symbolic mentions” (ibid, 2013: 156) of particular issues, that means that not all that is extrapolated from the conclusions has specific policy relevance. However, in both cases the dataset still appears to be especially useful to my purpose. Firstly, because the thesis will follow a “pure saliency approach” (ibid, 2013: 156), focusing on the number of mentions that the
issue of unemployment receives over time rather than the gravity of its discussion. Secondly, we shouldn’t be impaired by the sometimes-symbolic mention of the issue of unemployment in the conclusions, since the European Council is in most cases not in a position to introduce particular policy proposals.
Indeed, the dependent variable will track the number of mentions of unemployment related issues by exploiting two major codes in the dataset: 1:103 (specific mentions of the unemployment rate) and 5:500 (generic mentions of job creation and job growth) in each conclusion over the total number of mentions for that document; that is, the relative weight of unemployment concerns vis-à-vis all others. The choice of taking both objects into consideration is justified for two reasons: firstly, concerns with the unemployment rate alone may be too narrow. This is because the unemployment rate as a metric only becomes particularly salient when there are very steep changes in the average rate across the EU for most countries. This will likely only occur during times of international crisis and will be missing when facing more sectoral concerns for the unemployment rate. On the other hand, code 5:500 is more effective in this respect since, under close inspection, it marks the frequency of council mentions for the words “job creation” and “job growth”, which are intrinsically tied to the question of unemployment particularly for what concerns peripheral countries. The second reason why this choice is appropriate resides in the fact that being concerned with two items instead of one increases the overall variation in the object of inquiry, in this case the variable “Job concerns”. This will increase the number of observations
in the sample, thereby increasing the power of both tests and consequent estimations.
Public Opinion and the Eurobarometer
The first hypothesis expects the European Council agenda to be causally influenced by shifts in public opinion on job concerns over time. In order to correctly operationalise this variable, I resort to the most comprehensive and reliable source of survey data on EU level public opinion: the Eurobarometer. It was first introduced by the European Commission in 1973 surveying over samples of 1000 respondents assessing national priorities and perceptions on European integration. Nowadays, the Eurobarometer is one of the largest databases in the world and has been represented in several studies on democratic responsiveness at the EU level (see Wratil 2015; Alexandrova et al., 2016). In particular, I operationalise the attention devoted to job concerns by exploring the Most-important-problem (MIP) question in the European Union. The complexities deriving from using this approach are two-fold: to begin, data availability for this particular question is very limited. The MIP has been systematically recorded from 2003 onwards, meaning that the sample being observed is reduced compared to the complete dataset given by the Comparative Agenda Project. Secondly, the question is not asked consistently to respondents over time; formulations of the question, and nuances in the wording may have a direct effect on the respondent’s responses to the survey. This, however, is a typical downside of using survey data in empirical research, therefore we take the validity of the dataset to be granted by the series of substantive empirical findings
recorded by using this very same methodology. This will be operationalised as a time series vector at semester-level, identifying the proportion of respondents per country that raised unemployment as one of the three most important problems in the EU. However, to aggregate from nation-level to EU level data, the risk of taking a simple average would have been to underestimate EU level public opinion through the omission of population size as a determining factor. For statistical validity, this thesis uses a weighted average of these proportions based on year-specific country population sizes.
Measuring the ideological “Centre of Gravity”
In order to account for possible co-variational patterns between ideology and the Council, I resort to using the Comparative Manifesto Project (CMP) database, a digital resource consisting of coded electoral programs from 1945 up to 2018 (Lehman et al., 2018; Merz et al., 2016). The database covers mostly OECD and Eastern European countries, meaning it is highly effective for the purpose of indicating political positions for European countries. It is composed by the highly expert, native-language coders for each country, that have broken down each party manifesto prior to an election into a series of quasi-sentences with a unique topic or category. Indeed, this approach resorts to the well-known practice of inferring political positions from party manifestos rather than surveying individual voters to indirectly infer party positions on a left-right dimension (Budge et al., 1987). However, this method has received some criticisms, mainly concerning the adequacy of human coding instead of more advanced and modern machine learning applications, particularly for what concerns the
RILE predefined scale (Mikhaylov et al., 2012). Further still, some have put forward more theoretical cases against the CMP; Zulianello (2014) provides the argument that competition really should be measured in issues rather than policies, such that the CMP may be undermined by a theoretical, when not simply methodological, flaw. However, both concerns are largely overcome in this thesis. Firstly, the fact that machine learning applications may be more efficient than single coding procedures carried out by humans is largely irrelevant in the case at hand – I do not resort to using the RILE predefined scale which could be – or rather has been shown to be – controversial in various circumstances. On the other hand, I create a different scale that focuses solely on the pure saliency of jobs, salaries and trade unions. In this way, the measurement error is minimal when the coding is only dependent on the saliency of a very narrow range of topics and concepts weighing positively or negatively. The second concern is also overcome, since this thesis doesn’t want to focus on issue competition but on the saliency of job concerns as a policy objective. For example, employment policy is one of the main policies which a government is confronted with, but the issues through which this policy may be carried out are multiple. Policy-makers could choose to increase fiscal spending or reduce taxation. They could also attempt to raise import tariffs if and when particular industries were suffering from foreign competition. For our concerns, the issue that is dealt with by government is not as important as the prominence that creating jobs as a policy has on the party platform. This leads us to focus on the number of mentions, positive or negative, that jobs and salaries have for each party to
