Introduction
In the preceding chapters of this volume, we have developed a set of concepts and measurement tools to characterize policymaking and the nature of the policy debate in the wake of policy proposals put forward by governments in order to come to terms with the refugee crisis. In Chapter 1, we introduced the notion of politicization, which captures how salient and how polarized the given policy debate becomes among the political elite. In Chapters 6, 7, and 8, we focused on the intensity of the conflict between the respective actors in the debate. A crucial component underpinning these measures, which forms the backbone of our PPA dataset, is what we call the issue direction of the actions; in other words, whether the actor undertaking the action expresses a broad agreement or disagreement, or takes a neutral stance toward the policy in question. Aggregating these issue direction codes for a given unit of analysis – an entire episode, a given time period in an episode, or for a given actor – provides a glimpse of where the political elite (or particular elite groups) stand on the policy and by extension, how much elite resistance the governments face when enacting the policy.
In line with our previous analyses, we use the political elite in a rather holistic sense; not only does it capture the entire government apparatus as well as the parliamentary wings of the ruling coalition, but it also includes opposition parties (both the mainstream opposition and radical challengers), civil society organizations (including humanitarian groups, social movement organizations, experts, media, union, church, and other organized actors outside the party-political arena), and international actors (EU institutions and other governments). Such a broad interpretation of the political elite notwithstanding, we emphasize its distinction from the “demand side” of the policymaking equation that we focus on in Chapter 13: general public opinion in relation to the refugee crisis. In other words, the notion of elite support that constitutes the dependent variable of this chapter refers to the average position held by actors who act on behalf of political institutions or organizations with a capacity to reach and influence the opinion of broad audiences and the general public.
In an important departure from the previous chapters, the empirical analysis we provide in the present chapter takes an explicitly longitudinal perspective. Rather than asking why certain episodes face more or less elite support (on average), we inquire about the determinants of the ups and downs of such support over time within a given episode, while relegating much of the between-episode variation in support levels to episode-specific fixed effects. This is not to say that we consider context as irrelevant. Episodes play out across different geographical units in different time periods and in different issue domains. Accordingly, we will introduce some of the contextual variables – namely country type, episode type, and the phase of the refugee crisis – as possible moderators of the longitudinal drivers of elite support. Moreover, we distinguish between two types of drivers of support, which also serve as the most important organizing principle of this chapter: exogenous drivers that affect the overall level of elite support at any given point in time and endogenous interactions between actors. The analysis of the interactions investigates whether the ups and downs of support by one type of actor affect the contemporaneous or subsequent levels of support by another type of actor. The implicit theoretical framework we adopt thus assumes that in addition to the pressure of the crisis that affects all political actors involved in the policymaking process, actors also engage in strategic interactions, weighing the pros and cons of supporting government initiatives, voicing their opposition, or staying in the shadows of neutral ambiguity.
The unit of analysis of the chapter is the episode-month. With this choice, we aim to strike a balance between a temporal unit that is amenable to a meaningful aggregation of elite preferences (i.e., capturing enough observations for valid measurement), the availability of other longitudinal indicators as independent variables (e.g., problem pressure in the form of refugee flows and political pressure from the radical right are indicators that are available only on a monthly basis), and sufficient granularity that allows us to construct proper time series for statistical analysis. Especially the latter consideration proved somewhat problematic because fifteen of the forty episodes in our study lasted less than ten months, and ten episodes less than half a year. The prevalence of such short time series in our data is an important feature to keep in mind when we discuss some of the methodological choices in the empirical analysis. With this caveat in mind, the choice of the episode-month as the unit of analysis yields a time-series cross-section (TSCS) dataset of 644 observations with sufficient statistical power for valid inference.
The chapter is organized as follows. The next section provides a brief overview of the literature on elite support behind policymaking, emphasizing a lacuna present in the field: a heavy focus on the consequences of such support in the policymaking process for policy output and for public opinion in contrast to the relatively scant attention paid to its drivers. The third section aims to fill in this lacuna by putting forward a set of theoretical expectations on the drivers of elite support both on the systemic and on the actor-specific level. The fourth section describes the most important methodological choices, while the fifth section constitutes our empirical analysis, again structured along the systemic and the actor-specific levels. The last section concludes with a summary of our main findings and their implications.
The Importance of Elite Support behind Government Policies
In an abstract sense and from the perspective of various constraints that governments face, the policy challenge in the refugee crisis is no different from other policy challenges in the past. In addition to the exigencies of the underlying problem pressure, the constraints imposed by public opinion, and the institutional capacity of the state to address the problem, governments also need to reckon with potential dissent among the political elite. Alternatively put, the degree to which the government is able to rally elite support behind its policies may be a crucial determinant of the policies’ success.
The empirical literature on this matter documents a close link between the degree and type of elite support and policy output. One strand in this literature is largely informed by the American experience on asymmetric policy representation (Bartels Reference Bale2016; Hacker and Pierson Reference Hacker and Pierson2010). Gilens and Page (Reference Gingrich and Häusermann2014) take stock of a large dataset of policy issues (1,779 in total) and show that the final policy output bears a closer resemblance to business groups’ and the economic elites’ preferences than to those of the average citizens and mass-based interest groups. Grossmann, Mahmood, and Isaac (Reference Gruber, Barlai, Fähnrich, Griessler and Rhomberg2021) find a similar effect of organized groups in American politics. However, the authors emphasize partisan differences with regard to the type of organized interest whose support tends to determine the policies’ ultimate fate. Burstein and Linton (Reference Burstein and Linton2002) provide a meta-analysis from published political science journal articles to evaluate the relative importance of different kinds of elite support. They arrive at a more nuanced conclusion: Political organizations’ support for policies is most likely to have an impact when it resonates with the electoral concerns of politicians. That said, electoral considerations do not typically feature in most accounts of interest group politics. A case in point is Schamis (Reference Schattschneider1999), who emphasizes the role of elite support by business groups behind economic liberalization in Latin America, in contrast to the popular view of technocratic insulation at the top being the factor most conducive to structural reform programs (Haggard and Kaufman Reference Hatton2018). In neither of these accounts do electoral considerations take a central role, suggesting that elite support (or the lack thereof) has an autonomous influence on the policy process.
That said, elite groups’ impact on public opinion may very well be an alternative mechanism through which elites influence policymaking. Such an impact has been most extensively documented in the foreign policy domain via survey experiments. Whether respondents are cued by celebrity endorsements of a given course of action proposed by the government (Frizzell Reference Frizzell2011) or are exposed to the views of the military elite or policy advisors (Golby, Feaver, and Dropp Reference Gray, Lewicki, Gray and Elliott2018; Saunders Reference Scarpa and Schierup2018), their responses tend to align more closely with the proposed policy compared to nontreated individuals in these experiments. Similar effects were found in the issue domain of climate change policies. Kammermann and Dermont (Reference Kasimis and Kassimi2018) study the interaction between elite opinion and citizen preferences across a range of climate change policies in Switzerland and uncover the impact of elite support behind such policies on public opinion. Rinscheid, Pianta, and Weber (Reference Ripoll Servent and Trauner2021) come to similar conclusions in an empirical setting based on American respondents with regards to support for the phase-out of fossil fuel–powered cars and the deployment of carbon capture technology: When political parties endorse one of these policies, citizens’ support for the policies increases but only when they trust the party in question. All in all, elite support for public policies that are high on the political agenda across various issue domains and by various elite groups appears to have a clear and consistent link with public support for the issues in question. In other words, not only does elite support facilitate the enaction of public policies, but it also goes a long way in selling them to the public.
A common feature of these accounts is that implicitly, they tend to take elite preferences as given. While this assumption may be valid for certain types of elite groups in the case of certain types of policies (such as the role of business groups vis-à-vis economic liberalization or the role of military elites vis-à-vis foreign interventions), in the face of novel types of problem pressure with uncertain consequences, such as the refugee crisis we are studying, this assumption is highly problematic. Besides its impact on the policy process, we thus need to ask about the origins of elite support. In this regard, however, the extant literature provides significantly fewer cues. Though various features of elite groups and the policy debate, such as the institutional networks connecting elites (Van Gunten Reference Vincenzi2015), horizontal trust between them (Weinberg Reference Werts, Scheepers and Lubbers2022), or the role of policy framing when faced with policies that potentially conflict with their interests (Teigen and Karlsen Reference Tétényi, Barczikay and Szent‐Iványi2020) have been identified as possible determinants of elite support, we lack a coherent account of the origins of elite support behind government policies. Tellingly, in a two-wave survey of MEPs’ immigration attitudes some time before the refugee crisis, Lahav and Messina (Reference Laubenthal2005) document a convergence of views without specifying the driving mechanisms. This chapter thus takes up the task of specifying some of these mechanisms and subjects them to empirical testing on our PPA dataset.
Exogenous Drivers and Endogenous Interactions between the Elites in the Refugee Crisis
To begin conceptualizing the potential drivers of elite support for the forty national policy episodes that we study in this volume, we would like to remind the reader of the basic building blocks of our theoretical framework as outlined in Chapter 2. In one way or another, all the episodes were responses to the mounting problem pressures in the form of asylum applications overburdening the capacity of the countries’ asylum systems. Moreover, as policymakers sought to address the problem by erecting border fences and/or reforming the countries’ asylum system, they also had to reckon with pressures emanating from the exceptional salience of immigration in the minds of the public and the rising fortunes of radical right challenger parties that were uniquely well positioned to capitalize on the crisis. Throughout the book, we refer to these two forms of political constraints as political pressure.
How is the political elite expected to react in the face of such pressures? When these pressures mount, various elite groups are likely to weigh the expected costs and benefits of support and opposition to the governments’ policy initiatives. Though the salience of the refugee crisis is likely to vary across the different elite groups, they have a shared interest in putting the issue on the political agenda, lest they risk appearing out of touch with the concerns of ordinary citizens or a narrower subset thereof who are directly affected by immigration (e.g., via wage competition). Once they engage with the policies, elite groups next have to decide whether the expected cost of supporting the initiatives outweighs the expected benefits. The cost of support mainly derives from sharing responsibility for the potential failure of the policy in controlling refugee flows and/or equipping the asylum system for the reception and integration of asylum claimants. The benefits of support in turn derive from the perception that elite groups take the problem seriously and act as a voice of ordinary citizens clamoring for policy solutions. This consideration chimes in with the “rally-around-the-flag” perspective in crisis politics (Mueller Reference Müller and Rietig1970); In response to rising problem pressure and expectations from the general public, even potential dissenters among the political elite are under pressure to suspend criticism and temporarily support government initiatives. An additional source of benefits accruing from such support by the elite is that it allows them to come across as temporary policy allies of the government, which in turn may prompt the latter to offer policy concessions on other issue domains that are of greater importance for the respective elite groups.
The balance of these considerations thus suggests that in response to the crisis shock, the elite is expected to provide a temporary boost to government initiatives. However, we need to disentangle the impact of the two separate sources of pressures. While problem pressure may indeed prompt the elite to align behind governments, political pressure – especially when it comes from the radical right – may act as a countervailing force. Political pressure from the radical right is a signal to the other elite groups that public discontent with the proposed policies is palpable, and acting as a voice of such discontent may thus become a viable political strategy. This consideration leads us to put forward the baseline hypothesis for this chapter that expects opposite impacts emanating from the two sources of pressure that governments face.
H1: While mounting problem pressure leads to a (temporary) boost to elite support, increasing political pressure prompts the elite to oppose the proposed policy initiatives.
These considerations, however, need to be contextualized. Mounting problem and political pressures have vastly different implications across the types of countries, the types of episodes on the agenda, and the different phases of the refugee crisis. The underlying conditioning principle across these contexts is the notion of contestability. We posit that the degree to which the political elite perceives the government initiatives as contestable depends on their country’s structural location vis-à-vis refugee flows, the range of possible policy alternatives on the table, and the availability of a precedent and policy templates to be borrowed from other countries.
Starting with country types, an important distinction lies between the strategic calculation of elite actors in frontline countries on the one hand and transit and destination countries on the other. In frontline countries, the authorities can credibly scapegoat EU institutions and other member states for failing to relieve the asymmetrical burden that these countries face solely due to the “bad luck” of geography. In such a context, openly opposing government initiatives to get the problem under control carries the risk of being seen as obstructing the national cause and contributing to the perceived sense of injustice and grievances. In transit and destination countries, on the other hand, the appropriate policy response to the crisis is more contentious, pitting proponents of a relatively open asylum regime against the hardliners clamoring for a closed-door immigration regime in general and an uncompromising stance on sticking to the Dublin rules in particular. In this context, the public positions of elite groups are likely to diverge, giving rise to a higher level of dissent to government initiatives compared to the frontline states.
Similar considerations apply for the differential response of the elites across episode types. Again, elite support in response to mounting pressure is likely to depend on the perceived viability of policy alternatives. These alternatives are more likely to be present in cases of asylum reform because eligibility criteria, appeal conditions, return procedures, and various other aspects of the asylum systems are subject to legitimate contention in the absence of an acute sense of urgency to act. Border controls, by contrast, are desperate moves to get the situation under control with no other immediate policy alternative being in sight. In the face of such an emergency, it is thus considerably riskier for elite groups to challenge governments. We thus expect that elite incentives to dissent in the face of rising problem and political pressure are reduced when border control measures are on the political agenda.
Finally, we expect the temporal evolution of the broader refugee crisis to act as a third moderator of the elite response to government initiatives. As we outlined in Chapter 5, the refugee crisis can be conceptualized in terms of three distinct phases. The first phase is characterized by rising refugee flows across the Mediterranean without any overarching European or even national response commensurate to the scale of the problem to come. The second phase begins in the early summer of 2015 with the first border control measures imposed by the Hungarian authorities and the first European push toward a communitarian solution (the Relocation Scheme). The peak (and end) of the second phase is the EU–Turkey agreement signed in March 2016. We regard the period following the agreement as a distinct phase because with the externalization of border controls, refugee flows were significantly reduced, and the sense of urgency considerably abated. We argue that elite incentives to respond to the problem and political pressures vary across the phases. The greatest scope for dissent exists in the early phase, when no European or national policy solutions are forthcoming as templates that governments can adopt, and critical voices against early policy initiatives come across as highly credible in the absence of viable templates. As countries put up border barriers one after the other in the heat of the crisis and remolded their asylum systems in a comparable fashion, these critical voices became less credible, and the pressure on the elite to fall in line increased. We thus expect that in the later phases, elite groups have become more likely to support government initiatives in response to rising problem and political pressures. To summarize the three conditional hypotheses below:
H2a: Rising problem and political pressures lead to a higher level of dissent among the political elite in transit and destination states than in frontline states.
H2b: Rising problem and political pressures lead to a higher level of dissent among the political elite during policy debates on asylum rules compared to debates on border control measures.
H2c: Rising problem and political pressures lead to a higher level of dissent among the political elite during the early phase of the crisis compared to subsequent phases.
Thus far, we have implicitly treated the political elite somewhat monolithically, under the assumption that there is a common core of incentives they react to in a comparable fashion. We now relax the assumption and formulate expectations on group-specific behavior. Specifically, we identify four types of elite groups in line with the categorization we have put forward in Chapter 6. One part of the elite is closely affiliated with the government as members of governing parties, members of the cabinet, or members of other institutions under the direct control of the national government. We shall refer to these elites as governing elites. The second elite group we analyze separately consists of members of opposition parties, either from the mainstream opposition or from radical challengers. The third elite group consists of EU and other supranational institutions as well as foreign governments. Fourth, we also include in the analysis what we have broadly referred to as civil society groups, comprising social movement organizations, churches, unions, media actors, experts, academics, and other actors whose elite standing derives less from holding the levers of political power than from their potential to sway public opinion. We shall refer to these groups as civil society elites.
Though group-specific elite behavior may also depend on the external pressure that the governments face in the refugee crisis, we shall focus in this second part of the analysis on endogenous dynamics between elite groups, namely on their responsiveness to the actions of other elite groups who may be allies or potential rivals. An intuitive conceptualization of such responsiveness is the expected level of support for the governments’ policy initiatives by one elite group as a function of the changing levels of support provided by the other elite groups.
Starting with the governing elite, we expect that it is the most sensitive to opposition support because other elite groups have an indirect influence at best on the fate of the policy and on the electoral standing of governments. Opposition parties, by contrast, can present policy alternatives in parliament and other institutional venues, and the government is under pressure to react to such alternatives. Moreover, opposition groups have the potential to mobilize crowds, putting the government under pressure in the protest arena. However, it is an open question whether the government, in response to dissent from the opposition, closes ranks behind the proposal or whether, alternatively, critical voices within the government are reinforced and it splits on the issue, especially when the opposition strategically seeks to precipitate such splits (Whitaker and Martin Reference Wihtol de Wenden2022). Given the urgency of the problem pressure the government faces and the electoral threat from the radical right, we expect closing ranks to be a more likely scenario because, under heightened media scrutiny, any split is likely to become public and detrimental to government survival, as exemplified by the splits in the Swedish governing coalition during the refugee crisis.
Turning to the opposition, it is most likely to respond to the actions of its potential allies. One such group of allies are critical voices in the government itself. However, splits in the government are likely to provide only temporary momentum to opposition offensives because potential dissenters within the government are unlikely to want to risk losing office by providing open support to opposition parties. International and civil society support, by contrast, are more reliable power resources because they have the potential to legitimate opposition discourse. Empirically, the opposition and civil society groups have been shown to act in concert against government proposals on various occasions in the recent past (Kriesi et al. Reference Bojar, Gessler, Hutter and Kriesi2020). We thus expect that lower (higher) levels of support behind government policies by international and civil society groups are likely to decrease (increase) support by the opposition.
Does a mirror logic apply for the determinants of support by international institutions and civil society? To some extent, the likely answer is affirmative: Both international and civil society groups have something to gain when they wish to express opposition to government policies and other elite groups share their critical stance – because coordinated opposition is likely to legitimate dissent. However, we expect this logic to hold particularly for civil society groups because international actors need to be seen as (at least somewhat) neutral arbiters between the governing elite and dissenting groups. It is particularly risky for international actors to openly side with opposition forces, lest they be accused of interfering with domestic politics of a sovereign member state. We thus expect that civil society groups are likely to voice dissent in response to dissent by the opposition and international actors, whereas the position of international actors is less likely to be driven by the position of the governments’ domestic opponents. To summarize these expectations in a third set of hypotheses below:
H3a: The governing elite is most responsive to opposition dissent. Specifically, in response to dissent (reduced levels of support) by the opposition, the governing elite is likely to close ranks and reduce dissent (increase levels of support) within its own ranks.
H3b: Opposition groups are likely to increase dissent (reduce levels of support) in response to dissent by civil society groups and international actors.
H3c: International and civil society groups are likely to increase dissent (reduce levels of support) in response to dissent by civil society groups and international actors, respectively.
H3d: Civil society groups are more likely to increase dissent (reduce levels of support) in response to dissent by the opposition rather than international actors.
Method: A Longitudinal Analysis of Elite Support
As already indicated in the introduction to this chapter, our sample consists of a total of 644 observations where the unit of analysis is the episode month. In theory, the data structure is well suited for a TSCS (time-series cross-section) design with episode fixed effects to control for the systematically different average levels of support across units (episodes) that may be correlated with the models’ covariates. What sets the dataset apart from the most common TSCS designs is that the episodes (or at least a subset thereof) do not unfold simultaneously, and there is a considerable imbalance in the length of the individual series. One serious complication that arises from the relatively short (T < 10) series for a large part of the episodes is the well-known dynamic panel data bias (Nickell Reference Niemann and Speyer1981) in case of a dynamic specification. To assess the gravity of the problem, we inspected the dynamic nature of the dependent variable, the average level of elite support both in its systemic and in its actor-specific forms.Footnote 1 Reassuringly, the dependent variable displays little persistence over time, with an autoregressive coefficient of around 0.2. To visually convey this lack of persistence of the series, Figure 10.1 shows the evolution of the dependent variable for the ten longest episodes in our sample. The sudden spikes and drops of the series indicate that shocks dissipate rather quickly, making the behavior of the dependent variable not all that dissimilar from a white noise series. Substantively speaking, this pattern is somewhat puzzling at first because one would expect relatively stable elite preferences toward a given policy initiative over time. However, one must remember that the support variable is an aggregated measure of various actors and depending on which particular institutions act in a given episode month, it is likely to be rather volatile. Moreover, while some episode-months are rich in action, others are averages of only a handful of observations, further adding to observed volatility. With this volatility in mind, we consider that specifying the models in static terms poses little risk for biased coefficient estimates and goes a long way toward addressing the dynamic panel data bias in the case of short time series.
Figure 10.1 The evolution of average elite support over time
Other complications that may arise in time-series cross-section designs is the biased estimates for the standard errors of the coefficients due to panel-level heteroscedasticity and autocorrelation in the residuals. To get around this problem, we employ Beck and Katz’s (Reference Beck and Katz1995) recommended tool of panel corrected standard errors with a Prais–Winsten correction of the residuals. As the autocorrelation coefficient estimates (rho) reveal, however, autocorrelation was a marginal concern in most of the models, further underscoring the white noise–like behavior of the dependent variable.
All the estimated models shown below thus regress the average level of elite support (either in its systemic or its actor-specific form) on the key covariates of interest, a control for an episode-specific time trend and the episode-specific fixed effects. The key covariates in the first part of the analysis are the pressure indicators, all standardized between 0 and 1 so that the coefficient estimates are directly comparable on the same scale. The key covariates in the second part where we seek to predict the group-specific levels of support are the support variables for the other three elite groups. In the baseline models that test for the overall impact of the exogenous drivers (H1) as well as the models testing for interactions between the actors (H3a–H3c), contextual variables are included in the models as controls. In the models that test the three conditional hypotheses (H2a–H2c), interactions of the conditioning characteristics (type of state, episode, and phase) with the pressure indicators are introduced as additional covariates.
One important and open-ended decision we had to take was the temporal form of the time-varying covariates (exogenous drivers and support levels of the other actors). We had no strong theoretical priors to inform us whether the covariates should be introduced in a contemporaneous form or with lag(s). Introducing them with lags has the advantage of guarding against simultaneity bias, especially in the case of the interactive models, where the different elite groups may influence each other in a reciprocal fashion. However, not allowing for contemporaneous impacts runs the risk of arriving at false negative conclusions based on the coefficient estimates because a month may be a long enough time for a change in elite behavior to show up in the policy debate. The pragmatic compromise we took was running separate models for contemporaneous impacts and for one- and two-month lags.Footnote 2 In case of multiple coefficients showing up as significant, we selected as the final model the one that produced the best fit based on the R2 statistic.
Results
We begin the analysis with laying out the baseline model in Table 10.1. The passage of time (see “counter” variable) exerts a small but steady drop in average levels of elite support, suggesting that elites tend to distance themselves from the policy proposals over time. The impact over a month is a trivial-sounding 0.007, but over a year it accumulates to over 0.08 (with the dependent variable defined on the –1 to 1 scale). In terms of country types, destination countries tend to display a significantly lower level of elite support, in accordance with our expectation regarding the more contentious nature of policymaking in such countries. The impact is rather large, amounting to a 0.6 lower level of support in these countries compared to frontline states. Furthermore, the third phase of the crisis that begins with the signing of the EU–Turkey agreement tends to be associated with higher levels of elite support. This is again consistent with the idea that with the availability of policy templates from other countries, the scope and incentives for elite dissent are reduced. Finally, the estimates for the policy type (a dummy for asylum reforms as distinct from border control measures) are also largely consistent with our expectations that the scope for dissent is greater in the case of asylum policies, even if the effect is not significant.
Table 10.1 The impact of problem pressure and political pressure on levels of support behind government policies
| Model I: baseline | Model II: country types | Model III: episode types | Model IV: periods | |
|---|---|---|---|---|
| L2.problempressure | –0.402 | 9.943 | –0.577 | –6.839 |
| (1.93) | (2.05)* | (2.55)* | (3.15)** | |
| L.politicalpressure | –0.871 | 1.985 | –1.184 | –0.831 |
| (2.63)** | (1.90) | (2.26)* | (1.20) | |
| Mid-crisis | 0.153 | 0.139 | 0.152 | 0.142 |
| (1.84) | (1.68) | (1.84) | (0.59) | |
| Late-crisis | 0.217 | 0.186 | 0.219 | 0.146 |
| (2.22)* | (1.89) | (2.23)* | (0.58) | |
| Transit | 0.176 | 1.726 | 0.188 | 0.252 |
| (1.68) | (1.81) | (1.73) | (1.97)* | |
| Destination | –0.561 | 2.046 | –0.701 | –0.610 |
| (3.56)*** | (2.49)* | (2.87)** | (3.65)*** | |
| Asylum | –0.127 | 0.599 | –0.682 | –0.027 |
| (0.88) | (0.60) | (1.66) | (0.13) | |
| Transit*L2.problempressure | –10.342 | |||
| (2.12)* | ||||
| Destination*L2.problempressure | –10.416 | |||
| (2.14)* | ||||
| Transit*L.politicalpressure | –1.684 | |||
| (1.35) | ||||
| Destination*L.politicalpressure | –3.522 | |||
| (3.11)** | ||||
| Asylum*L2.problempressure | 0.440 | |||
| (1.03) | ||||
| Asylum*L.politicalpressure | 0.492 | |||
| (0.75) | ||||
| Mid-crisis*L2.problempressure | 6.398 | |||
| (3.02)** | ||||
| Late-crisis*L2.problempressure | 5.680 | |||
| (2.48)* | ||||
| Mid-crisis*L.politicalpressure | –0.389 | |||
| (0.63) | ||||
| Late-crisis*L.politicalpressure | –0.175 | |||
| (0.27) | ||||
| Counter | –0.007 | –0.008 | –0.008 | –0.007 |
| (2.38)* | (2.62)** | (2.57)* | (2.35)* | |
| Constant | 0.731 | –1.630 | 0.975 | 0.919 |
| (2.71)** | (1.98)* | (2.40)* | (2.58)** | |
| N | 38 | 38 | 38 | 38 |
| n | 566 | 566 | 566 | 566 |
| R2 | 0.26 | 0.28 | 0.27 | 0.28 |
| Rho | –0.02 | –0.03 | –0.02 | –0.03 |
* p < 0.05; ** p < 0.01; *** p < 0.001
The central question of this chapter, however, concerns the reaction of the elites to rising problem and political pressure. The coefficient estimates of the corresponding coefficients are only partly in line with our expectations. Rising problem pressure is associated with a significantly lower level of elite support two months later, though the estimate is just short of the 5 percent significance level when the variable is introduced together with the political pressure variable. Therefore, in contrast to the rally-around-the-flag dynamics, if anything, the elite appears to distance themselves from the government initiatives in response to mounting problem pressure. We can only speculate at this point about the driving mechanism behind this effect. However, it appears to be the case that highlighting the potential risks (or outright failure) of the proposed policy remedies is viewed by the elite as the less risky option compared to tagging along with the governments’ agenda. The estimate is also substantively meaningful: The predicted difference between elite support between the sample minimum and the sample maximum of the standardized problem pressure variable is 0.40. The impact of political pressure, by contrast, is in the expected negative direction, with a substantively larger impact compared to problem pressure: The difference between the sample minimum and the sample maximum of political pressure gives rise to a 0.90 difference in the predicted level of elite support. Finally, no statistically significant effect is found for the impact of political pressure emanating from the higher salience of the immigration issue, so we omitted this coefficient estimate from the final models.
Before we proceed to the interactive models, it is worth stressing the difference in the temporal dynamics between the two pressure variables. While the impact of problem pressure shows up with two-month lag, political pressure exerts an instantaneous and one-month lagged impact on elite behavior (though we included only the one-month lag in the final model for ease of interpretation). One possible explanation is that rising political pressure is not just a trigger but also a manifestation of elite discontent. In other words, as the elite turns away from governments, a part of the electorate takes note by turning toward parties that own the immigration issue in general and act as the loudest critics of governments’ handling of the refugee crisis in particular.
How do these impacts vary across the contextual characteristics we have identified above following the notion of contestability? We reestimate the models with each pressure variable introduced in interaction with the contextual covariates (Models II, III, and IV), and we calculate the marginal effects (Figures 10.2 and 10.3) of the pressure variables across the different contexts. Similar to the baseline models, the salience variable does not produce any significant estimates across any of the contexts. The impact of problem pressure, however, clearly separates frontline states from transit and destination states. In fact, the negative overall estimate we have found for the problem pressure variable (at its second lag) is restricted to transit and destination states, whereas in frontline states, we find a large positive estimate (amounting to a change in average level of support of around 10 for a full swing between the sample minimum and sample maximum level of problem pressure). This impact clearly lies outside the range of the elite support variable, but this is due to the fact that the typical level of problem pressure – as measured by submitted asylum claims – is considerably lower in frontline states than in transit and destination states. In fact, the sample maximum in this country group on the standardized scale of the problem pressure variable is a mere 0.03, so the estimated positive impact needs to be evaluated accordingly: Moving from the sample minimum to the sample mean in this country group, for instance, amounts to a change of 0.3 on the scale of the average support variable.
Figure 10.2 The impact of problem pressure across country types, episode types, and crisis periods
Figure 10.3 The impact of political pressure across country types, episode types, and crisis periods
Turning to the conditional impact of problem pressure across episode types, there is some evidence for the conditioning role of episode types, but the impact goes against our expectations. Rather than asylum reforms, it is during debates on border measures that rising problem pressure leads to a higher level of dissent by the political elite, as is evidenced by the negative and significant estimate of the problem pressure variable during such episodes. The conditional role of the crisis periods, however, conforms to our expectations: While the impact of rising problem pressure in the early phase leads to a large drop in elite support, the impact is small and statistically indistinguishable from 0 in the subsequent phases. To quantify the impact, we need to again consider that the typical level of elite support was considerably lower in this first phase of the crisis. A move from the sample minimum to the sample mean in this period (0.03 on the standardized problem pressure scale) thus amounts to a drop of around 0.2 in elite support.
The conditioning role of political pressure from the radical right across the contextual characteristics we study in this chapter is very similar to that of problem pressure. While the political elite in destination states reacts to rising political pressure by stepping up dissent, the corresponding estimates in transit and frontline states are statistically indistinguishable from 0. As for episode types, we observe a pattern identical to the impact of problem pressure: Contrary to expectations, it is border control debates that prompt the elite to oppose policy initiatives in response to political pressure, whereas the impact of this form of pressure during asylum debates is nonsignificant. Finally, in contrast to the impact of problem pressure, the impact of political pressure does not appear to significantly diverge across the phases of the crisis. The point estimates for political pressure throughout all the crisis phases are negative and comparable in size, though the estimate falls short of significance in the first crisis phase, probably due to the relatively few episodes and observations falling in this phase of the crisis. As for the substantive size of the estimates, most of the impacts are larger than for problem pressure. For instance, a move from the sample minimum to the sample maximum political pressure in destination states amounts to a drop of 0.8 in elite support. The corresponding move in border episodes is even greater, amounting to a drop of no less than 1.2 in elite support. Overall, in line with the baseline models, we can claim that the substantive impact of political pressure is larger than that of problem pressure.
Turning to the impact of interactions between the elites, Table 10.2 shows the model estimates with group-specific level of support as the dependent variable and the contemporaneous and the lagged levels of support by the other elite groups as the key independent variables alongside the contextual controls and episode-specific time trends.
Table 10.2 Actor-specific models predicting levels of support for government policies
| Model I: government | Model II: opposition | Model III: international | Model IV: civil society | |
|---|---|---|---|---|
| Opposition support | 0.142 | |||
| (3.65)*** | ||||
| L2.opposition support | –0.130 | |||
| (2.87)** | ||||
| International support | –0.106 | 0.107 | ||
| (1.97)* | (2.14)* | |||
| Civil society support | 0.127 | 0.069 | ||
| (2.80)** | (2.17)* | |||
| Asylum | –0.192 | –0.068 | –0.159 | –0.188 |
| (4.20)*** | (0.18) | (1.56) | (1.81) | |
| Transit | –0.382 | 0.748 | –0.380 | –0.194 |
| (2.92)** | (1.87) | (2.10)* | (2.60)** | |
| Destination | –0.350 | 0.151 | –0.269 | 0.034 |
| (9.00)*** | (0.80) | (2.19)* | (0.75) | |
| Trend | –0.006 | 0.001 | –0.002 | 0.001 |
| (2.03)* | (0.30) | (1.16) | (0.52) | |
| Mid-crisis | 0.277 | 0.024 | 0.031 | –0.136 |
| (3.21)** | (0.34) | (0.69) | (2.35)* | |
| Late-crisis | 0.267 | –0.076 | 0.076 | –0.045 |
| (2.53)* | (0.88) | (1.26) | (0.59) | |
| Constant | 0.569 | –0.556 | 0.383 | 0.048 |
| (5.32)*** | (1.39) | (2.22)* | (0.59) | |
| N | 38 | 39 | 40 | 40 |
| n | 566 | 605 | 644 | 644 |
| R2 | 0.23 | 0.30 | 0.26 | 0.27 |
| Rho | 0.05 | 0 | 0.04 | 0.02 |
| Fixed effects | YES | YES | YES | YES |
* p < 0.05; ** p < 0.01; *** p < 0.001
Starting with the government itself (Model I), its level of support only appears to be influenced by opposition dissent, in line with our expectations. The impact is significant only at its second lag. Dissent by the opposition thus appears to push potential dissenters within the government to fall in line. Alternatively put, at higher levels of opposition support, government dissenters are under less pressure to close ranks and feel freer to express reservations about the government’s policy initiatives.
Model II provides only partial evidence for the legitimating momentum that we expected to play a role behind opposition behavior. Though the impact of civil society support is positive and significant in its contemporaneous form, there is a simultaneous negative impact of international support. Rather than gaining legitimacy from international actors’ criticism of the government, a tentative interpretation of this finding is that the opposition is under pressure to line up behind governments when the latter are under attack from international actors. Regardless of the particular mechanism at play, it seems that domestic civil society elites are more reliable allies of opposition parties when it comes to decisions to oppose the governments’ policy initiatives.
In terms of the last two actor types (Models III and IV), our expectations regarding the more limited impact on international actors’ behavior compared to civil society elites’ behavior are well supported by the data. Support by international actors appears to be independent of both government and opposition support, in line with our expectations that they cannot be seen as openly taking sides in domestic political conflict. Civil society elites’ support, however, does affect international actors’ behavior. Though the estimate is substantively smaller than the estimates found for government and opposition actors’ behavior, it nevertheless suggests that international actors are emboldened in their criticism of national governments when notionally independent domestic groups step up their own criticism of the policy initiatives. Finally, the largest and most consistent estimates are found for the behavior of these civil society groups: The direction of their support follows the change in support of international and opposition groups. The legitimation logic that we expected to drive the behavior of the three elite groups outside the governing elite is thus borne out most clearly for civil society groups by the data.
When thinking of these interaction patterns among elite groups that we have uncovered in the group-specific longitudinal analysis, a cautionary note is in order. By allowing for contemporaneous estimates due to the possibility of relatively quick reactions (within a month window) that may not show up in the lagged estimates, we opened up the possibility of simultaneity bias. The possibility of such simultaneous causation is especially pertinent when the reversal of the dependent and independent variables in the respective models produces similar estimates. In our models above, opposition–civil society interactions are a case in point. The impact of civil society support on opposition behavior and the impact of opposition support on civil society behavior are both estimated to play out simultaneously at a comparable magnitude. This may indeed be a sign of mutually legitimating dynamics between the respective parties, but more advanced longitudinal techniques, such as vector autoregressive (VAR) models, would be needed to disentangle the particular causal order among the elite groups’ reaction pattern.
Conclusion
In this chapter, we sought to introduce a longitudinal perspective in the study of the policy debates of the refugee crisis at the national level. Specifically, we aimed to uncover the determinants of elite support – broadly understood – behind government policies in the context of the forty policy episodes that we study throughout the book. We have seen that somewhat surprisingly (partly due to the heterogenous nature of elite groups), the average level of support by the elite shows considerable volatility over the course of the policy episodes. We conjectured that some of this temporal fluctuation can be explained by three different sets of variables: the changing political and problem pressure that governments face, the contextual characteristics that may moderate this relationship, and the endogenous dynamics unfolding between different elite groups.
Though many of these drivers indeed turn out to be statistically significant and substantively important drivers of elite support, some of the patterns we have found partly or fully went against our prior expectations. Thus, far from the elite closing ranks behind government proposals as the “rally-around-the-flag” perspective may suggest, nongovernment elites rather use the strategic opportunity offered by mounting problem pressure to articulate opposition to these proposals and signal distance from governments as a result. However, this dynamic is mostly confined to destination and transit states, and it is more prominent during debates on border controls and in the early phase of the crisis. By contrast, the impact of political pressure is largely in line with our expectations: In response to the growing strength of the radical right, the elite steps up dissent, with the strongest effect found, again, in destination states. A tentative explanation for why elites are particularly sensitive to these pressures in destination states is that these governments had the highest “degree of freedom” as far as the management of the crisis is concerned; hence, they proved the most vulnerable to domestic political conflict when the risk of policy failure became manifest.
In addition to responding to external pressure, elite groups were also shown to engage in strategic behavior with respect to each other. While dissenters within governments are responsive only to partisan opposition actors, the behavioral calculus among opposition, civil society, and international actors is more complex. In one way or another and to different degrees, they follow in each other’s footsteps and form a latent alliance against government proposals. An exception to this rule is the opposition’s reaction to international intervention: In response to criticism from international actors, opposition parties tend to side with governments, arguably in response to an increasingly critical public opinion of the EU’s and the international community’s management of the refugee crisis.
These strategic responses of various elite groups to each other add an important insight to one of our previous chapters (Chapter 6) on domestic conflict lines. We showed in that chapter that the bulk of the conflict played out between governments and (some of) their domestic and international opponents depending on a host of contextual characteristics of the episodes. What remained hidden in that analysis due to the lack of a longitudinal dimension is how these opponents dynamically interact. The inclusion of such a longitudinal dimension allowed us to shed light on this omission: The governments’ opponents systematically respond to each other’s expressed level of support to the government’s initiatives, albeit sometimes with substantial lags. Though the government, by virtue of its central role in the policy process, is indeed the main originator or the target of conflict, other actors hardly act in isolation when they decide on their response strategies.
An important limitation of this elite-focused analysis is its disproportionate focus on the supply side of the policy process. Though the inclusion of our two political pressure variables did incorporate public opinion as a potential driver of elite behavior, our dataset did not provide sufficiently rich and systematic information on the most visible and audible voices of public engagement: protest activity.
Open access