Scholarly article on topic 'Do parties matter in delegation? Partisan preferences and the creation of regulatory agencies in Europe'

Do parties matter in delegation? Partisan preferences and the creation of regulatory agencies in Europe Academic research paper on "Political Science"

0
0
Share paper
Academic journal
Regulation & Governance
OECD Field of science
Keywords
{""}

Academic research paper on topic "Do parties matter in delegation? Partisan preferences and the creation of regulatory agencies in Europe"

Regulation & \

Governa nee

Regulation & Governance (2014) ••-•• doi:10.1111/rego.12072

Do parties matter in delegation? Partisan preferences and the creation of regulatory agencies in Europe

Laurenz Ennser-Jedenastik

Institute of Political Science, Leiden University, Leiden, The Netherlands Department of Government, University of Vienna, Vienna, Austria

Abstract

The ideological orientation of parties in government has not been prominently featured in explaining the rise of regulatory agencies. This paper argues that theories based on political uncertainty and credible commitment can yield meaningful predictions regarding the relationship between government preferences and the establishment of regulatory agencies, when ideological orientation is linked with notions of party competence and issue ownership. The empirical section tests three such hypotheses with data on the establishment of 110 regulatory agencies in 20 European democracies between 1980 and 2009, thus providing one of the most comprehensive cross-national analyses of agency creation to date. The results show that ideologically extreme cabinets are more likely to establish regulatory agencies and that right-wing governments create more agencies in the economic than in the social domain. These findings partly qualify the view on the scarce relevance of government preferences in explaining the rise of the agency model in regulation and that the emulation mechanism of the diffusion process is the dominant force behind agencification.

Keywords: delegation, ideology, parties, regulatory agencies.

1. Introduction

During the past decades, the state has retreated from the direct provision of services in many areas (e.g. telecommunications and electricity), to overseeing the functioning of markets dominated by private corporations. Regulation has replaced direct engagement as the dominant relationship between government and private actors in many sectors of the economy. This "rise of the regulatory state" (Majone 1994) has led to the creation of a large number of regulatory agencies (RAs) that often operate at a considerably higher level of independence from elected officials than the core bureaucracy (Gilardi & Maggetti 2011). RAs have not only been established in the economic realm, but also for consumer protection, environmental safety, and health care, thus contributing to a profound transformation in the organization of the public sector across a variety of policy domains throughout Europe and beyond (Levi-Faur 2003,2005, 2006; Jordana et al. 2011; Verhoest et al. 2012; Kleibl 2013).

Correspondence: Laurenz Ennser-Jedenastik, Department of Government, University of Vienna, Rooseveltplatz 3/1/126b, 1090 Vienna, Austria. Email: laurenz.ennser@univie.ac.at Accepted for publication 24 August 2014.

The copyright line for this article was changed on 13 January 2016 after original online publication. © 2014 Blackwell Publishing Asia Pty Ltd.

This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.

Given the comprehensive nature of this development, it is worth examining under what conditions politicians are willing to delegate authority over regulatory policy to an agency. The most prominent explanations revolve around the notions of credibility (Gilardi 2002, 2005a, 2008; Wonka & Rittberger 2010), political uncertainty (Moe 1990a, b; Moe & Wilson 1994; de Figueiredo 2002), diffusion (Gilardi 2005b; Levi-Faur & Jordana 2006; Jordana et al. 2011), and external pressures, such as Europeanization (Gilardi 2005b, 2008).

Although ideology1 has been found to play a major part in moves towards privatization and liberalization that often precede (and cause) regulatory reform (Boix 1997; Bortolotti et al. 2001, 2004; Zohlnhofer et al. 2008; Belloc &Nicita 2011,2012; Obinger et al. 2014), the creation of RAs has not typically been linked to politicians' ideological leanings - at least not in Europe. Whereas the regulatory state and its institutions in the United States emerged as a product of the progressive era (Glaeser & Shleifer 2003), the development towards "regulatory capitalism" in Europe and elsewhere (Levi-Faur 2005) has usually been regarded as a technocratic response to liberalization, privatization, and increased pressures of globalization.

As Levi-Faur notes, regulatory reform does not square easily with conventional ideological preconceptions. While "at the ideological level neoliberalism promotes deregulation, at the practical level it promotes, or at least is accompanied by, regulation" (Levi-Faur 2005, p. 14). This insight has contributed to the view that "[c]ommentators have generally rejected ideological explanations of regulatory reform" (Vogel 1996, p. 21).

This paper re-examines ideology's supposed marginal role in explaining the creation of regulatory agencies. Yet rather than providing a genuine theory of parties' role in agency creation, it combines theoretical arguments from research on regulation with concepts from the literature on party competition. The theoretical section, thus, focuses on the credibility and the political uncertainty theses and discusses their implications once we take into account that parties (and, thus, governments) vary in their ideological complexion, policy emphasis, and perceived issue competence. Three hypotheses emerging from this discussion are examined with original data on the establishment of 110 regulatory agencies in seven policy domains across 20 European countries between 1980 and 2009, thus extending on the empirical design proposed by Gilardi (2005b, 2008). The results show that ideologically extreme governments are more likely to establish agencies and that right-wing governments are more likely to create regulators in the economic than in the social domain.

2. Theoretical approaches to the rise of regulatory agencies

Multiple theoretical arguments have attempted to explain the widespread adoption of the agency model across countries and regulatory domains. The following paragraphs cover two of the most prominent approaches - credibility and political uncertainty - and highlight their implications for government preferences in the creation of regulatory agencies. To this end, theories of agency creation are linked to the literature on party competence and issue ownership.

Other approaches to agency creation mostly adopt a diffusion perspective (Gilardi 2005b; Levi-Faur & Jordana 2006; Jordana et al. 2011), arguing that the agency model is transferred within and across national and sectoral channels. While the diffusion perspective is clearly important, it is not central to the analysis presented here and will, therefore, be featured only as a control.

This paper does not propose a new framework for theorizing about the role of parties and governments in the creation of regulatory agencies, but rather tries to look at existing arguments

in the agency literature and connect them logically to well-established concepts in party research. The basic argument underlying the following discussion is the idea that political uncertainty and the need for credible commitment do not apply uniformly across all types of political parties and all policy domains. It conjectures that the incentives to insulate policy choices or the need for credible commitment vary systematically with government preferences and policy domains. This assumption gives rise to a number of hypotheses that can be tested by examining variation in the timing of agency creation across countries and regulatory sectors.

2.1. Political uncertainty

Political uncertainty is an inherent feature of any democratic system of government. No democratically constituted governments' survival is guaranteed beyond the next election. This implies not only that the incumbent government may at some point be fully or partly replaced by its competitors, it also means that tomorrow's government may have preferences other than those of today's (Moe 1990a, 1991; Potoski 1999; de Figueiredo 2002). Politicians who anticipate such a potential shift in preferences have strong incentives to find ways of protecting their policy choices, making them more difficult to undo in the future. As Gilardi (2008, p. 60) argues, the delegation of powers to regulatory agencies can be a means to insulate certain policies by removing them from the direct control of elected politicians. While it is not impossible for future governments to overturn such decisions, the establishment of an agency creates an institutional barrier to sweeping and immediate policy shifts.

The notion that political uncertainty and the desire to insulate policies drive agency creation (leading to high levels of agency independence) is not only well established in the theoretical literature. Substantial empirical evidence has also proven its validity (Gilardi 2005b, 2008; Wonka & Rittberger 2010). Not all studies, however, can find significant effects (Elgie & McMenamin 2005; Elgie 2006). Yet what are the implications of this argument for the role of parties and government preferences?

First, because political uncertainty is typically conceived as the risk of being replaced by an ideologically distant alternative government (Franzese 2002), it could be argued that there is a curvilinear relationship between political uncertainty and government ideology. Therefore, all else equal, centrist governments and parties are less likely to suffer massive policy losses when they are replaced, whereas more extreme2 actors stand to lose a great deal. Overall, we would, therefore, expect more extreme governments to be more likely to resort to agency establishment as an insulation strategy than moderate ones.

A prime example can be found by looking at Margaret Thatcher's three cabinets during the 1980s and 1990s. These happen to be the most right-wing cabinets in the dataset that created a number of agencies (e.g. in telecommunications, energy, financial markets and medicines, but also in a number of domains not covered in this study). As Vogel (1996, p. 131) argues, "Thatcher administration officials favored independent regulators because of the dynamics of alternance in British politics." Given the huge ideological divide between Labour and the Conservatives in the 1980s, it was all the more important for the Conservatives to institutionally protect their reform policies against reversal by a future Labour cabinet.

H1. Ideologically extreme governments are more likely to establish regulatory agencies

than moderate ones.

Furthermore, we can assume that policy-oriented parties with different ideological leanings vary not only in their specific preferences but also in their policy priorities. In other words, some parties care about certain policy domains more than others do (Budge & Farlie 1983;

Green-Pedersen 2007). In addition, voters perceive parties to be more competent in handling these "owned" issues (Petrocik 1996; Van der Brug 2004; Walgrave et al. 2012). Typically, social and welfare policy are owned by social democrats (Blomqvist and Green-Pedersen 2004). Environmental issues are generally associated with Green parties (Meguid 2005, 2008), while immigration is the domain of the populist radical right (Ivarsflaten 2008), and economic or fiscal policy are often viewed as the "home turf" of conservative and liberal (in the European sense) parties (Conradt & Dalton 1988; Meyer & Müller 2013).

Theories of party and voter behavior have routinely drawn on these assumptions (Budge & Farlie 1983; Petrocik 1996; Budge 2001; Van der Brug 2004; Dolezal et al. 2014). They are strongly supported by the finding that multiparty governments distribute ministerial portfolios according to the issue priorities that can be ascribed to specific party families (Budge & Keman 1990; Bäck et al. 2011).

Yet the assumption that parties differ in how they prioritize issues does not only inform studies of electoral competition and government formation; it also overlaps with work on policymaking. The policy output generated by governments is in many ways a function of parties' emphasis on certain issues (Hofferbert & Klingemann 1990; Klingemann et al. 1994; Knill et al. 2010; Egan 2013). It has even been shown that party issue emphasis can have an impact on agency design (Bertelli 2006).

The realization that parties place distinct value on different policy domains also has implications for the role of political uncertainty. The fact that some policy areas are more central than others influences a party's calculus in anticipation of possible defeat. It suggests that the losses incurred by a government relegated to the opposition are not uniformly distributed, but rather they vary with the importance ascribed by government parties to each policy area. The anticipated policy damage is greatest in the most highly prioritized domains. As a consequence, the incentive for incumbents to resort to insulation strategies is greatest in exactly those areas. As agencies are created to protect specific policies from interference by future governments, these creations are a function of incumbents' issue priorities.

H2. Governments are more likely to establish regulatory agencies in their most highly

prioritized domains.

2.2. Credible commitment

The ability of governments to credibly commit to a specific course of action in the future is a highly valuable asset. As elaborated by Shepsle (1991) there are two basic ways to establish credibility. A commitment can be motivationally credible, meaning that the government's preferences at t+1 will, in fact, be in line with its commitment at t. In such cases, commitment is incentive-compatible and, thus, self-enforcing. External actors have every reason to believe that the government will not deviate from its stated course of action if the government's preferences so dictate.

However, in other cases it may be that governments have time-inconsistent preferences. Thus, politicians who announce some future policy at time t may prefer a different policy at time t+1. External actors anticipating such temporal inconsistency will adapt their behavior not to the stated policy, but to their expectations about the policy at t+1, thus undermining the government's ability to effectively set policy. In the face of potential for divergence between government preferences at t and t+1, a commitment can be made imperatively credible by means of a commitment device, such as coercion by other actors (e.g. the judiciary) or delegation of powers over the implementation of the policy.

The credibility problem has been most prominently discussed in the literature on monetary policymaking (Rogoff 1985; Cukierman 1992; Keefer & Stasavage 2003). Electoral incentives make it difficult for governments to credibly commit to a non-expansionary monetary policy in the future. As Election Day approaches, government politicians will often find it tempting to loosen monetary policy in order to stimulate output and employment, and, thus, renege on earlier promises to keep inflation low. Rational actors (e.g. wage bargainers) will anticipate this temporal inconsistency in the government's preferences and adapt their expectations accordingly. This leads to higher (and, thus, suboptimal) inflation levels even if governments intend to stick to a low-inflation policy.

A typical solution to this problem is to give authority over monetary issues to a conservative central banker who "does not share the social objective function, but instead places 'too large' a weight on inflation-rate stabilization relative to employment stabilization" (Rogoff 1985, p. 1169). Thus, monetary policy is made credible by delegating powers to an agent who: (i) has different preferences, and (ii) possesses the power to enforce them. As Majone (2001) notes, this logic may also extend to regulation and helps explain why the top personnel in regulatory agencies are often chosen not only for their expertise, but also for their commitment to a specific policy agenda that may diverge considerably from that of the government or the median voter.

Indeed, the importance of credibly committing to a policy goal has long been recognized by students of delegation and regulatory politics (Levy & Spiller 1994,1996; Spiller 1993). The idea that independent regulatory agencies can serve as devices to back up policy commitments by politicians has also been widely applied in the literature (Majone 1997; Gilardi 2002; Thatcher 2002, pp. 130-1; Elgie & McMenamin 2005; Wonka & Rittberger 2010). The most obvious example is in utilities regulation. A credible commitment not to give preferential treatment to individual firms (e.g. former state monopolists after privatization) can be an important mechanism to encourage investment and market entry by outside competitors (Gilardi 2008, pp. 32-5). The creation of an RA backs up the commitment to non-interventionism and, thus, enables competition by guaranteeing a level playing field, even if the government's preferences change in a more interventionist direction. In addition, credible commitment may also help to rebuild consumer confidence after environmental or food scandals, especially if the government is considered close to the perpetrators.

The credibility thesis is often examined not by looking at agency creation (measured dichotomously: establishment or non-establishment). A more fine-grained approach entails observing the level of independence granted to regulators (Gilardi 2005a; Maggetti 2007; Wonka & Rittberger 2010; Hanretty & Koop 2012). Yet because institutions are sticky and levels of agency independence are not likely to change frequently, analyzing the timing of agency establishment with event history models allows for a better grasp on the impact of time-varying determinants, such as government preferences. Also, formal independence only goes so far as a commitment device because the de-facto independence of an agency may not correspond to its level of legal autonomy.

How, then, does the credibility thesis relate to arguments about government ideology? Here, Shepsle's (1991) distinction between motivationally and imperatively credible commitments comes into play. As the creation of regulatory agencies is an institutional mechanism to produce imperatively credible commitments, it is more likely to be employed in cases where motivation-ally credible commitments cannot be made. Shepsle's argument, however, implies that the level of policy credibility is a function of the government's preferences. Put simply, right-wing governments can make claims in some policy areas with greater credibility than left-wing

governments (and vice versa). Therefore, a government's "credibility deficit" will vary with its ideology and across policy domains.

This idea is hardly novel. Extant research has shown that fiscal consolidation is more successful if the type of adjustment goes against the conventional wisdom about government ideology and spending preferences, thus signaling credible commitment on the governments' part (Tavares 2004). Others have demonstrated that interest rates can, in fact, increase under left-wing governments, but only in the presence of an independent central bank (Belke & Potrafke 2012). Left-wing governments seeking to maintain low inflation, thus depend to a much larger extent on independent central banks to compensate for the credibility deficit that they face with respect to tight monetary policy.

Furthermore, Gilardi (2008, pp. 114-7) finds that left-wing governments are more likely to establish utility regulators after liberalization than right-wing governments who have greater motivational credibility to ensure pro-market regulation and, thus, less need to resort to an institutional commitment device. The third hypothesis applies this logic on a more general level.

H3. Governments are more likely to create regulatory agencies in domains in which their

policy commitments are motivationally less credible.

2.3. Translating the hypotheses into testable predictions

Whereas the first hypothesis (related to the impact of ideological extremeness) is relatively straightforward in its empirical implementation, the other two require more specification. The best possible test of H2 and H3 would rely on cross-national data on issue ownership and party competence across a range of regulatory domains. The lack of such data requires relying on some assumptions and proxy measures.

To this end, the analysis will: (i) map governments along a one-dimensional left-right axis, and (ii) draw on a basic distinction between economic and social regulation (Gilardi 2005b, 2008; Wonka & Rittberger 2010). Economic regulation is primarily concerned with establishing well-functioning markets, promoting competition, and preventing dominant corporations from abusing their market power. This is especially so in newly liberalized sectors previously controlled by state monopolies. By contrast, social regulation is typically understood as aiming "to protect people or the environment from the damaging consequences of industrialization" (Hawkins & Hutter 1993, p. 199).

The assumption that links this distinction to government preferences is that, in line with issue ownership theories, governments and parties of different ideological complexions vary in their policy priorities and their perceived competences. Left-wing parties put more weight (and are considered more credible) on social issues, welfare, employees' rights, consumer interests, or the environment, whereas right-wing parties are more concerned with (and perceived as more credible on) economic issues, well-functioning markets, and competition.

Importantly, this is not to say that left-wing parties are against competition or that right-wing parties do not care about consumers or the environment. The argument is simply that different ideological tendencies lead to differences in policy emphasis, perceived competence, and credibility.

Because of these underlying assumptions, H2 and H3 become competing hypotheses in the empirical analysis. The hypothesis based on the political uncertainty argument (H2) implies that left-wing governments put greater weight on issues covered by social regulation and are, therefore, keener on insulating policies by creating agencies in the social domain, whereas right-wing governments should establish regulators in the economic domain because economic policy features more prominently in their policy calculus. By contrast, the credibility argument

(captured in H3) suggests that left-wing governments have greater credibility deficits in the economic domain and should, therefore, create economic regulators, whereas right-wing governments are perceived as less credible in the social domain and should, thus, establish agencies there.

3. Data and method

Following Gilardi (2005b, 2008), the empirical strategy uses event history analysis with temporally-varying covariates to trace the effect of government ideology on the establishment of independent regulatory agencies in four economic (electricity, telecommunications, financial markets, and competition) and three social (food safety, medicines, and the environment) policy domains. The dependent variable is the time it takes for an agency to be established in a certain domain.

The analysis extends Gilardi's dataset by including a longer time series (starting in 1980 and leading up to 2009) and more countries (EU-15 plus Switzerland, Iceland, Norway, Malta, and Cyprus). Observations are country-domain-years, with right-censoring applied in all years except when the respective agency was created (subsequent observations are dropped). As many agencies go through splits, mergers, reforms, and renaming over the years, it is sometimes difficult to define when an authority was established. As a general rule, the first creation by law of an organizationally independent entity in the respective policy sector was taken as the relevant date. In a small number of cases, this leads to divergence from Gilardi's data (e.g. for the Finnish Food Safety Authority whose creation Gilardi dates to 2001, but whose predecessor was already established in 1990). In sum, agencies were established in 110 cases during the period of observation in 123 country-domains.3 The distribution of duration times until agency creation is displayed in Figure 1.

In order to gauge the effect of government preferences, a continuous and one-dimensional measure is created from expert survey data on party positions. It calculates the left-right position

0 5 10 15 20 25 30

Duration in years

Figure 1 Distribution of durations.

Note: N = 123 country-sectors, including 13 censored cases; mean = 17.1, median = 17, standard deviation = 6.8.

of all parties represented in a cabinet, weighted by their parliamentary seat share. Data on left-right positions are taken from a number of expert surveys conducted during the past three decades (Castles & Mair 1984; Laver & Hunt 1992; Benoit & Laver 2006; Steenbergen & Marks 2007; Hooghe et al. 2010). The values for the individual observations were selected so as to minimize the time span between the data points and the times at which the expert surveys were conducted. Also, the scale was calibrated to a range from 0 (left) to 10 (right). Each country-year was given the ideology score of the cabinet in office on the last day of the year, except when an agency was created in that year. In this case, the left-right score of the government in office at the time of creation (taken to be the date of the parliamentary vote to establish the agency) was used.

Two caveats come with the use of these data. First, expert surveys of party positions are not flawless (Budge 2000). Yet they typically provide more consistent and valid estimates than available alternatives, such as data generated from party manifestos, which have been criticized on theoretical and methodological grounds (Benoit & Laver 2007; Hansen 2008; Benoit et al. 2009; Mikhaylov et al. 2012; Dolezal et al. 2014).

Second, the ideologies of parties and governments cannot necessarily be mapped onto a single dimension. The left-right scale is a simplification over more complex models that use, for instance, an economic and a cultural dimension to represent party preferences. However, the available expert survey data do not provide consistent two-dimensional measures across time and for a large number of countries. Limitations in data coverage leave little choice but to use the one-dimensional left-right scale.

The key independent variables generated to operationalize the three hypotheses are, thus, the absolute left-right distance from the scale midpoint (H1),4 the simple left-right indicator, and its interaction with a social regulation indicator (H2 and H3).

A number of control variables are specified, again largely following Gilardi (2005b, 2008). First, an updated version of the Political Constraints Index (Henisz 2002) is used to account for overall levels of policy stability. Furthermore, the analysis replicates Gilardi's measures for Europeanization (operationalized as indicators for three specific EU directives), liberalization, and privatization in the telecommunications and electricity sectors (extending on the information provided in Gilardi 2008, p. 152)5 and the predictor for the diffusion of agencies. This is operationalized as the share of countries that have created an agency in a specific sector.

Also, a measure of the average duration (in years) of the three preceding cabinets is included. This captures an aspect of political uncertainty that is not accounted for by ideological extremeness. It serves as a proxy for political instability (and, thus, the likelihood of cabinet turnover) in a political system. On average, Italy scores lowest on this measure, whereas Switzerland and Luxembourg have the highest values.

A descriptive account of the independent variables is given in Table 1.

There may be many other unobserved factors that influence the establishment of RAs. In order to account for such variation, shared frailty terms are specified at the country level. Shared frailty terms account for intra-group correlation and are, thus, functionally equivalent to random effects (Box-Steffensmeier & Jones 2004, pp. 146-7; Cleves et al. 2010, pp. 156-64). Alternative specifications with country-fixed effects yield almost identical results (see Appendix).

The duration times are modeled with semi-parametric Cox proportional hazards regressions (Cox 1972). The Cox model has come to be seen as superior to fully parametric alternatives because it demands none of the theoretical assumptions regarding the distribution of duration times that are necessary for parametric models, such as the Weibull, Gompertz, or exponential regression. The Cox model has become the standard in many political science applications, yet

Table 1 Descriptive statistics of independent variables

Variable Description N Mean SD Min. Max.

Cabinet ideology Left-right score (0-10) 2220 5.06 1.59 1.05 7.8

Cabinet extremeness Absolute left-right deviation from scale 2220 1.33 0.86 0 3.95

midpoint (5)

Cabinet ideology X Interaction of left-right score with social 2220 2.45 2.79 0 7.80

social regulation regulation indicator

Social regulation Indicator for social regulation (0/1) 2220 0.48 0.50 0 1

Political constraints Updated version of the political constraints 2220 44.86 13.17 12.00 72.00

index (III) by Henisz (2002)

Average cabinet Average duration of last three cabinets in 2220 2.41 1.11 0.17 5.10

duration years

Liberalization Indicator for years when liberalization took 2220 0.01 0.09 0 1

place (only telecom and energy)

Privatization Indicator for years when privatization took 2220 0.01 0.08 0 1

place (only telecom and energy)

Diffusion (logged) Share of regulators created across countries 2220 3.19 0.62 2.30 4.59

by domain

EU Directive 92/44 Indicator for years 1992/93 in telecom sector 2220 0.01 0.11 0 1

EU Directive 97/51 Indicator for years 1997/98 in telecom sector 2220 0.00 0.06 0 1

EU Directive 96/92 Indicator for years 1996-99 in energy sector 2220 0.03 0.16 0 1

Several observations drop because of missing information on party positions for the caretaker cabinet Dini I (Italy) in 1995. The diffusion variable has been logged to meet the normality assumption. The political constraints index has been multiplied by hundred to make the interpretation of the multivariate results more straightforward. The years for the European Union (EU) directives were coded according to the time frames within which member states had to comply. Max, maximum; min, minimum; SD, standard deviation.

it requires that hazards are proportional over time. Tests based on scaled Schoenfeld residuals (Schoenfeld 1982; Grambsch & Therneau 1994) reveal that the political constraints covariate and the diffusion predictor violates this assumption in some model specifications. The standard remedy to this violation is to include an interaction term between the offending variable and some function of time (typically the natural log) into the estimation (Box-Steffensmeier & Jones 2004, p. 136).

4. Analysis

Before moving to the multivariate analysis, Table 2 provides a bivariate breakdown of government ideology and agency establishment in the economic and social sector. To that end, all observations were classified into terciles according to the respective governments' left-right score. The table shows that right-wing governments created more agencies overall (44), but this difference emerges only from the economic domain. Left-wing governments created an almost equal number of economic and social regulators (55% vs. 45%), whereas three quarters of all regulators set up by right-wing governments were in the economic realm. Centrist governments fall somewhere in between. Table 2 indicates that, while governments of all persuasions have created regulatory agencies in all domains, there is a tendency in the data that is in line with H2.

Table 2 Cabinet ideology and agency creation in economic and social regulation

Left-wing tercile Centrist tercile Right-wing tercile Total

Economic regulation N 18 20 33 71

% 55% 61% 75% 65%

Social regulation N 15 13 11 39

% 45% 39% 25% 35%

Total N 33 33 44 110

% 100% 100% 100% 100%

Note: Even though the data suggest a link between the two variables, the relationship is not statistically significant (Cramer's V = 0.18, p-value = 0.15). However, this table only looks at instances of agency creation, while the multivariate models also include data spells for years in which no agency was created. A bivariate analysis of agency creation and ideology including all annual data spells yields significant correlation coefficients.

To see whether this pattern holds in a multivariate environment, event history models are employed in the analysis. Cox proportional hazards regressions with shared frailties specify the hazard rate for each observation i in country j as

hj (t ) = he (t )a, ej )

where h0 is the unspecified baseline hazard, a is the frailty parameter (with mean 1 and variance d), x is a set of covariates, and / is a vector of regression coefficients. Table 3 presents four regression models. First, the impact of government ideology is examined as a baseline scenario. Model 2 introduces the extremeness measure to test H1. Model 3 shows that the effect of extremeness persists even when excluding the original ideology predictor. Model 4 includes all covariates.6

The hazard ratios reported in Table 3 have a relatively straightforward interpretation. A hazard ratio of 1.2, for instance, would indicate a 20 percent increase in the probability of an agency being established as the respective covariate rises by one. A number smaller than one would mean that an agency is less likely to be created with each increase in the independent variable.

Model 1 shows that the left-right predictor alone has no impact on agency creation. The hazard ratio of 1.1 is in line with the descriptive account given in Table 2 (more agencies are created by right-wing governments), but at P = 0.16, the effect is not statistically significant. This suggests that there is no uniform impact of ideology on the likelihood of agency creation, which conforms to the findings reported by Gilardi (2008, p. 114).

However, once the extremeness measure is introduced in Model 2, the picture changes. The significant hazard ratio of 1.37 suggests that moving one point from the center of the left-right scale towards the extremes increases the likelihood of agency creation by about 37 percent. In other words, the relationship between ideology and agency establishment is not linear. In order to better evaluate this finding, Figure 2 plots the inverted survival curves (which in this context can be interpreted as the cumulative probabilities of agency creation over time) for three different government types (left, right, and centrist). All other variables are held constant at their respective means (continuous variables) or modes (dichotomous variables).

It is clearly visible in the graph that the data conform to H1. Both left-wing and right-wing governments display higher probabilities of agency creation over time than moderate govern-

Table 3 Determinants of agency establishment

Model 1 Model 2 Model 3 Model 4

Cabinet ideology 1.106 1.076 1.192**

(1.42) (1.07) (2.05)

Cabinet extremeness 1.370** 1.407** 1.390**

(2.04) (2.23) (2.12)

Cabinet ideology X social regulation 0.757** (-2.11)

Social regulation 1.696 (0.65)

Political constraints 0.905** 0.918* 0.914* 0.928

(-1.98) (-1.65) (-1.72) (-1.41)

Political constraints X ln (time) 1.037** 1.033* 1.035* 1.028

(2.04) (1.78) (1.84) (1.49)

Average cabinet duration 1.369 0.954 0.880 0.974

(0.44) (-0.06) (-0.17) (-0.03)

Average cabinet duration X ln (time) 0.827 0.962 0.989 0.968

(-0.72) (-0.14) (-0.04) (-0.12)

Liberalization 2.353* 2.466* 2.452* 2.195

(1.67) (1.76) (1.75) (1.52)

Privatization 1.984 1.643 1.484 1.860

(1.12) (0.77) (0.61) (1.00)

Diffusion (by sector) 0.000* 0.000* 0.000* 0.948

(-1.77) (-1.75) (-1.75) (-0.04)

Diffusion (by sector) X ln (time) 72.645** 68.898** 68.868**

(2.27) (2.26) (2.26)

EU directive 92/44 0.948 0.996 1.025 0.813

(-0.07) (-0.00) (0.03) (-0.25)

EU directive 97/51 10.809*** 9.496** 8.970** 11.349***

(2.66) (2.48) (2.41) (2.68)

EU directive 96/92 0.709 0.726 0.723 0.828

(-0.70) (-0.65) (-0.66) (-0.38)

N 2220 2220 2220 2220

N (country-sectors) 123 123 123 123

N (agencies created) 110 110 110 110

Log likelihood -428.1 -429.0 -429.6 -422.7

0 (estimated frailty variance) 0.240*** 0.25*** 0.27*** 0.30***

Note: Figures are hazard ratios from Cox proportional hazard regressions with shared frailties by country; t-statistics in parentheses; *P < 0.1, **P < 0.05, ***P < 0.01. EU, European Union.

ments. In the figure, the step curves for the non-centrist governments intersect the 50 percent threshold after 19 (right) and 22 (left) years, whereas centrist cabinets reach that point only after 27 years. The first hypothesis is, thus, supported by the data. Centrist cabinets are less likely to create agencies than governments whose ideological center of gravity is located at a more extreme position.

Model 4 includes an interaction of cabinet ideology with a dichotomous indicator for social regulation. Both the original variable and the interaction term are significant. The ideology predictor yields a hazard ratio of 1.19, which can be interpreted as the effect for cases in which

Figure 2 The effect of cabinet extremeness on agency creation.

Note: Depicted values are inverted survival functions, calculated based on Model 2. All other covariates held at their respective means (continuous variables) or modes (indicator variables).

Economic regulation

Social regulation

Right-wing cabinet (7) Left-wing cabinet (3)

5 10 15 20 25 Time until agency established (years)

Right-wing cabinet (7) Left-wing cabinet (3)

5 10 15 20 25 Time until agency established (years)

Figure 3 Probability of agency establishment by ideology and regulatory type.

Note: Depicted values are inverted survival functions, calculated based on Model 3. All other covariates held at their respective means (continuous variables) or modes (indicator variables).

the social regulation variable takes on the value of zero. In other words, governments that are further to the right are more likely to create regulatory agencies in the economic domains.

To get a visual impression of these findings, Figure 3 depicts the inverted survival curves as a function of government ideology and the type of regulation (economic or social), with all other variables held constant at their respective means (continuous variables) or modes (dichotomous

variables). The black step function represents the cumulative probability of agency creation under a right-wing government (located at 7 on the ideological scale). The grey line denotes the same probability under a left-wing government (located at a left-right position of 3).

Note that after fifteen years (the midpoint of the 30-year period of observation), the probability that a regulator has been established is between 20 and 25 percent for left-wing cabinets (at a left-right position of 3), irrespective of the policy domain. However, under right-wing cabinets (at a score of 7), the probability of an RA having been created is 50 percent in the economic sector, but only 15 percent in the social sector. Ideological differences between governments thus become much more visible in the realm of economic regulation. The two grey lines take on a very similar trajectory, whereas the black step curve is much steeper in the left-hand panel of Figure 3.

This result provides an important clarification of the results from the regression models. Left-wing governments are not necessarily more inclined to establish social than economic regulators, but they are more likely to create RAs in the social sector and much less likely to establish economic regulators than right-wing governments. However, the two very similar curves for left-wing governments in Figure 3 should be interpreted against the background that the data cover many more creations of economic (71) than social regulators (39).

In sum, these findings provide solid support for H2, whereas H3 must be rejected. The effect of cabinet ideology on agency creation seems to conform to the logic that right-wing governments care more about economic regulation than left-wing governments and therefore establish more regulatory agencies in these domains.

A cursory look at the control variables in Table 2 suggests that average cabinet duration (another proxy for political uncertainty) has very little explanatory power. The interactions of the political constraints predictor with the log of time in Models 1 to 3 (included to mitigate the violation of the proportional hazards assumption) suggest that the effect of political constraints, a modified version of the concept of veto players (Tsebelis 1995, 2002), is negative at first, thus indicating lower probabilities of agency establishment. After just a few years, however, the effect becomes insignificant and remains indistinguishable from zero over the remaining period of observation (see Appendix for graph). This suggests that the overall effect of the political constraints index is very small, if present at all, thus corroborating the results reported by Gilardi (2008, 114).7

Liberalization only displays significant effects in the models excluding the social regulation indicator, and no impact at all in the country-fixed effects models (see Appendix). As this variable varies only within two of the economic policy domains (telecommunications and electricity), it is logical that there is a negative correlation with social regulation. This is most likely what causes the liberalization effect to disappear in Models 3 and 4. The privatization variable has no significant impact in any of the models. However, it is quite plausible that, as Gilardi (2008, p. 114) reports, the effects of these two variables are conditional on other covariates.

The diffusion predictor is only significant when the interaction with the log of time is also included. As shown in the Appendix, the effect is slightly negative at first (though only barely significant), then indistinguishable from zero, and becomes positive after about 14 years. One interpretation of this finding is that the effect of diffusion has become stronger over time. However, these results should be read with great caution. Because no agency in the sample is ever abolished without replacement (although quite a few of them are reformed), the diffusion measure as operationalized here (and in other studies, see Gilardi 2005b; Jordana et al. 2011) is necessarily an almost perfect correlate of time. This makes it problematic to include event history

models in which the time until occurrence of an event is the dependent variable. It is, thus, hardly surprising that it violates the proportional hazards assumption in many instances.

Finally, we observe a huge effect of one of the EU directives. Again, this result is consistent with Gilardi's analyses. All models also report d, the estimated frailty variance, as being different from zero, suggesting that important variation across countries is not captured by any of the independent variables in the model. Inspection of the country-fixed effects models reveals that Finland, Norway, Sweden, Ireland, and the United Kingdom are quicker to establish regulators than the reference country (Austria), whereas only Cyprus is slower. Such variation suggests that some administrative traditions (notably the Nordic and the Anglo-Saxon) have generally been more willing to embrace the agency model whereas others have been more reluctant. Yet as all of the models show, the conditional effect of ideology on regulatory agency creation holds despite such important cross-national variation.

Three central conclusions can be derived from the analysis. First, ideology can be asserted as a relevant driver of the establishment of regulatory agencies across countries and policy domains in Europe during the recent decades. The analysis is certainly limited in that it cannot explain the overall trend towards the adoption of the agency model in so many countries since the 1980s. However, it can contribute to our understanding of the micro-dynamics of this development, showing that ideology clearly conditions which type of government is more likely to establish which type of agency.

Second, the test of H2 against H3 clearly comes out in favor of the former. Governments do not necessarily establish agencies to bolster their credibility in policy areas where they do not have a "natural" competence advantage. Rather, it appears that governments create regulators in domains that they prioritize over others.

Third, a more general point can also be made. The main contribution of the analysis is to show that there is a systematic relationship between government preferences and the creation of RAs. Yet this relationship is not linear and varies across policy domains. There are also systematic differences between more extreme and more centrist governments.

As a caveat, it should be noted that the rejection of H3 does not amount to an overall refutation of the credibility thesis. One reason for this is that the operationalization of H3 is only one possibility to test H3. As outlined above, limitations in data availability clearly impair a perfect translation of the theoretical argument into an empirical analysis. In addition, there are many other ways in which empirical implications of the credibility thesis may become manifest. This is most evident in the high levels of formal independence that many RAs enjoy in areas where credible commitment is especially needed, such as utility regulation.

Also, the analysis here cannot cover all potential drivers of agency creation. External shocks, such as economic crises or environmental disasters, can have a strong impact on regulatory reform (e.g. many countries adapted their regulatory framework regarding food safety in the aftermath of the bovine spongiform encephalopathy crisis). Yet because it is unlikely that such events are strongly correlated with government ideology, the results presented above should hold even when taking external shocks into account.

5. Concluding remarks

The establishment of regulatory agencies across sectors and nations (and at the EU level) is one of the most relevant developments in the transformation of the state in Europe and beyond (Jordana et al. 2011). The analysis presented above adds to the large-N empirical accounts of agency establishment in Europe and offers a test of hypotheses drawn from theories based on

political uncertainty and credibility, enriched by concepts from research on political parties. The results show that more extreme cabinets are more likely to establish regulatory agencies, and that right-wing governments create more economic than social regulators. The first result is hypothesized to be a consequence of greater political uncertainty for extreme governments, as political uncertainty is higher if a government is at risk of being replaced by an ideologically distant alternative government (a risk that is invariably higher for non-centrist cabinets). The second finding supports the proposition that agency creation is - among other things - correlated with perceived party competence and/or party issue salience.

The analysis has, thus, contributed to our understanding of the role of government ideology in the institutional dynamics of the "rise of the regulatory state." The results provided by the event history models suggest that government preferences condition the trend toward the agency model as an institutional manifestation of regulatory politics. This secular increase in the number of independent regulatory agencies is part of a more general phenomenon of agencification, that is, an "apparent international convergence on the agency form" (Pollitt et al. 2001, p. 271; see also Van Thiel 2012, p. 22). Yet the findings presented above suggest that the secular increase at the aggregate level is very much a political process whose micro-dynamics are shaped by systematic variation across time, countries, and policy domains. While there has clearly been an overall trend towards the establishment of regulatory agencies, exactly which agencies are created when is also dependent on the ideological orientations of governments.

Furthermore, the fact that the delegation of powers to regulatory agencies is in part shaped by the ideological views of governing politicians highlights the potential of regulatory politics to become more strongly drawn into the arena of party competition. It also implies that political rationales can be influential for public sector reform in addition to explanations focused on the necessities of technical expertise and economic efficiency. Even though such pragmatic imperatives often appear to be paramount factors in regulatory reform, our understanding of the transformation of the public sector will be more comprehensive if we consider the role of ideological motivations.

Acknowledgment

The author gratefully acknowledges funding by the Austrian Science Fund (FWF) for the project 'Party Government, Patronage, and the Regulatory State', grant no. J 3409-G11.

Earlier versions of this paper were presented at the 3rd Annual Meeting of the European Political Science Association (EPSA) in Barcelona, June 2013, and at the MZES Colloquium at the University of Mannheim, September 2013. I would especially like to thank Martino Maggetti, Till Cordes, and Adam Chalmers for helpful comments.

1 Note that the terms "ideology" and "preferences" are used interchangeably throughout the paper.

2 I use the term "extreme" simply to denote actors that have policy preferences further removed from the center of the political space. In this sense, the term has no normative implications.

3 This number (123) is below the theoretical maximum of 140 (20 countries times 7 sectors) because several agencies were already established prior to 1980 and are thus excluded from the analysis.

4 Alternative specifications would be the absolute distance from the median or the mean, but because these figures are 4.92 and 5.06, the deviation from the scale midpoint (5) is practically equivalent.

5 Note that these variables take on the value of zero for all observations outside the domain of utilities regulation (i.e. financial markets, competition, medicines, food safety, and the environment).

6 Because the impact on agency creation is not obvious for a number of variables, alternative versions with one-year lags have been tested (for cabinet ideology, including its squared and interacted version, privatization, liberalization, and all EU directives). All of these alternative specifications considerably weakened the observed effects and most made them disappear. This suggests that the operationalizations chosen in Table 3 are the most appropriate.

7 Gilardi does find a significant effect of political constraints when interacted with liberalization. The direct effect, however, is insignificant.

References

Bäck H, Debus M, Dumont P (2011) Who Gets What in Coalition Governments? Predictors of Portfolio Allocation in Parliamentary Democracies. European Journal of Political Research 50, 441-478.

Belke A, Potrafke N (2012) Does Government Ideology Matter in Monetary Policy? A Panel Data Analysis for OECD Countries. Journal of International Money and Finance 31, 1126-1139.

Belloc F, Nicita A (2011) The Political Determinants of Liberalization: Do Ideological Cleavages Still Matter? International Review of Economics 58, 121-145.

Belloc F, Nicita A (2012) Partisan Privatizations: Assessing the Ideological Bias. Applied Economics Letters 19, 959-963.

Benoit K, Laver M (2006) Party Policy in Modern Democracies. Routledge, London.

Benoit K, Laver M (2007) Estimating Party Policy Positions: Comparing Expert Surveys and Hand-coded Content Analysis. Electoral Studies 26, 90-107.

Benoit K, Laver M, Mikhaylov S (2009) Treating Words as Data with Error: Uncertainty in Text Statements of Policy Positions. American Journal of Political Science 53, 495-513.

Bertelli A (2006) The Role of Political Ideology in the Structural Design of New Governance Agencies. Public Administration Review 66, 583-595.

Blomqvist P, Green-Pedersen C (2004) Defeat at Home? Issue-Ownership and Social Democratic Support in Scandinavia. Government and Opposition 39, 587-613.

Boix C (1997) Privatizing the Public Business Sector in the Eighties: Economic Performance, Partisan Responses and Divided Governments. British Journal of Political Science 27, 473-496.

Bortolotti B, Fantini M, Siniscalco D (2001) Privatisation: Politics, Institutions, and Financial Markets. Emerging Markets Review 2, 109-137.

Bortolotti B, Fantini M, Siniscalco D (2004) Privatisation Around the World: Evidence from Panel Data. Journal ofPublic Economics 88, 305-332.

Box-Steffensmeier JM, Jones BS (2004) Event History Modeling: A Guide for Social Scientists. CUP, New York.

Budge I (2000) Expert Judgements of Party Policy Positions: Uses and Limitations in Political Research. European Journal of Political Research 37, 103-113.

Budge I (2001) Theory and Measurement of Party Policy Positions. In: Budge I, Klingemann H-D, Volkens A, Bara J, Tanenbaum E (eds) Mapping Policy Preferences: Estimates for Parties, Electors, and Governments 1945-1998, pp. 75-90. OUP, Oxford.

Budge I, Farlie DJ (1983) Explaining and Predicting Elections: Issue Effects and Party Strategies in Twenty-Three Democracies. Allen & Unwin, London.

Budge I, Keman H (1990) Parties and Democracy: Coalition Formation and Government Functioning in Twenty States. OUP, Oxford.

Castles FG, Mair P (1984) Left-right Political Scales: Some 'Expert" Judgments. European Journal of Political Research 12, 73-88.

Cleves MA, Gould WW, Gutierrez RG (2010) An Introduction to Survival Analysis Using Stata 3rd Edition. Stata Corp, College Station, TX.

Conradt DP, Dalton RJ (1988) The West German Electorate and the Party System: Continuity and Change in the 1980s. The Review of Politics 50, 3-29.

Cox DR (1972) Regression Models and Life-tables. Journal of the Royal Statistical Society. Series B (Methodological) 34, 187-220.

Cukierman A (1992) Central Bank Strategy, Credibility, and Independence: Theory and Evidence. MIT Press, Cambridge, MA.

Dolezal M, Ennser-Jedenastik L, Müller WC, Winkler AK (2014) How Parties Compete for Votes: A Test of Saliency Theory. European Journal of Political Research 53, 57-76.

Egan PJ (2013) Partisan Priorities: How Issue Ownership Drives and Distorts American Politics. CUP, New York.

Elgie R (2006) Why Do Governments Delegate Authority to Quasi-autonomous Agencies? The Case of Independent Administrative Authorities in France. Governance 19, 207-227.

Elgie R, McMenamin I (2005) Credible Commitment, Political Uncertainty or Policy Complexity? Explaining Variations in the Independence of Non-majoritarian Institutions in France. British Journal of Political Science 35, 531-548.

de Figueiredo RJP (2002) Electoral Competition, Political Uncertainty, and Policy Insulation. American Political Science Review 96, 321-333.

Franzese RJ Jr (2002) Macroeconomic Policies of Developed Democracies. CUP, Cambridge, UK.

Gilardi F (2002) Policy Credibility and Delegation to Independent Regulatory Agencies: A Comparative Empirical Analysis. Journal ofEuropean Public Policy 9, 873-893.

Gilardi F (2005a) The Formal Independence of Regulators: A Comparison of 17 Countries and 7 Sectors. Swiss Political Science Review 11(4), 139-167.

Gilardi F (2005b) The Institutional Foundations of Regulatory Capitalism: The Diffusion of Independent Regulatory Agencies in Western Europe. The Annals of the American Academy of Political and Social Science 598, 84-101.

Gilardi F (2008) Delegation in the Regulatory State: Independent Regulatory Agencies in Western Europe. Edward Elgar, Cheltenham, UK.

Gilardi F, Maggetti M (2011) The Independence of Regulatory Authorities. In: Levi-Faur D (ed) Handbook on the Politics of Regulation, pp. 201-214. Edward Elgar, Cheltenham, UK.

Glaeser EL, Shleifer A (2003) The Rise of the Regulatory State. Journal of Economic Literature 41, 401-425.

Golub J, Steunenberg B (2007). How Time Affects EU Decision-making. European Union Politics 8, 555566.

Grambsch PM, Therneau TM (1994) Proportional Hazards Tests and Diagnostics Based on Weighted Residuals. Biometrika 81, 515-526.

Green-Pedersen C (2007) The Growing Importance of Issue Competition: The Changing Nature of Party Competition in Western Europe. Political Studies 55, 607-628.

Hanretty C, Koop C (2012) Measuring the Formal Independence of Regulatory Agencies. Journal of European Public Policy 19, 198-216.

Hansen ME (2008) Back to the Archives? A Critique of the Danish Part of the Manifesto Dataset. Scandinavian Political Studies 31, 201-216.

Hawkins K, Hutter BM (1993) The Response of Business to Social Regulation in England and Wales: An Enforcement Perspective. Law & Policy 15, 199-217.

Henisz WJ (2002) The Institutional Environment for Infrastructure Investment. Industrial and Corporate Change 11, 355-389.

Hofferbert RI, Klingemann H-D (1990) The Policy Impact of Party Programmes and Government Declarations in the Federal Republic of Germany. European Journal ofPolitical Research 18, 277-304.

Hooghe L, Bakker R, Brigevich A, De Vries C, Edwards E, Marks G et al. (2010) Reliability and Validity of the 2002 and 2006 Chapel Hill Expert Surveys on Party Positioning. European Journal of Political Research 49, 687-703.

Ivarsflaten E (2008) What Unites Right-wing Populists in Western Europe? Re-examining Grievance Mobilization Models in Seven Successful Cases. Comparative Political Studies 41, 3-23.

Jordana J, Levi-Faur D, Fernandez-i-Marin X (2011) The Global Diffusion of Regulatory Agencies: Channels of Transfer and Stages of Diffusion. Comparative Political Studies 44, 1343-1369.

Keefer P, Stasavage D (2003) The Limits of Delegation: Veto Players, Central Bank Independence, and the Credibility of Monetary Policy. American Political Science Review 97, 407-423.

Kleibl J (2013) The Politics of Financial Regulatory Agency Replacement. Journal of Politics 75, 552-566.

Klingemann H-D, Hofferbert RI, Budge I (1994) Parties, Policies, and Democracy. Westview Press, Boulder.

Knill C, Debus M, Heichel S (2010) Do Parties Matter in Internationalised Policy Areas? The Impact of Political Parties on Environmental Policy Outputs in 18 OECD Countries, 1970-2000. European Journal of Political Research 49, 301-336.

Laver M, Hunt WB (1992) Policy and Party Competition. Routledge, New York.

Levi-Faur D (2003) The Politics of Liberalisation: Privatisation and Regulation-for-competition in Europe's and Latin America's Telecoms and Electricity Industries. European Journal of Political Research 42, 705-740.

Levi-FaurD (2005) The Global Diffusion of Regulatory Capitalism. The Annals of the American Academy of Political and Social Science 598, 12-32.

Levi-Faur D (2006) Regulatory Capitalism: The Dynamics of Change beyond Telecoms and Electricity. Governance 19, 497-525.

Levi-Faur D, Jordana J (2006) Toward A Latin American Regulatory State? The Diffusion of Autonomous Regulatory Agencies Across Countries and Sectors. International Journal of Public Administration 29, 335-366.

Levy B, Spiller PT (1994) Institutional Foundations of Regulatory Commitment: A Comparative Analysis of Telecommunications Regulation. Journal of Law, Economics, & Organization 10, 201-246.

Levy B, Spiller PT (eds) (1996) Regulations, Institutions, and Commitment: Comparative Studies ofTelecom-munications. CUP, Cambridge, UK.

Maggetti M (2007) De Facto Independence After Delegation: A Fuzzy-set Analysis. Regulation & Governance 1, 271-294.

Majone G (1994) The Rise of the Regulatory State in Europe. West European Politics 17(3), 77-101.

Majone G (1997) Independent Agencies and the Delegation Problem: Theoretical and Normative Dimensions. In: Steuenberg B, Vught van F (eds) Political Institutions and Public Policy, pp. 139-156. Kluwer Academic Publishers, Dordrecht.

Majone G (2001) Nonmajoritarian Institutions and the Limits of Democratic Governance: A Political Transaction-cost Approach. Journal of Institutional and Theoretical Economics 157, 57-78.

Meguid BM (2005) Competition Between Unequals: The Role of Mainstream Party Strategy in Niche Party Success. American Political Science Review 99, 347-359.

Meguid BM (2008) Party Competition between Unequals: Strategies and Electoral Fortunes in Western Europe. CUP, Cambridge, UK.

Meyer TM, Müller WC (2013) The Issue Agenda, Party Competence and Popularity: An Empirical Analysis of Austria 1989-2004. Journal of Elections, Public Opinion & Parties 23, 484-500.

Mikhaylov S, Laver M, Benoit KR (2012) Coder Reliability and Misclassification in the Human Coding of Party Manifestos. Political Analysis 20, 78-91.

Moe TM (1990a) Political Institutions: The Neglected Side of the Story. Journal of Law, Economics, and Organization 6, 213-253.

Moe TM (1990b) The Politics of Structural Choice: Toward a Theory of Public Bureaucracy. In: Williamson OE (ed) Organization Theory: From Chester Barnard to the Present and Beyond, pp. 116-153. OUP, Oxford.

Moe TM (1991) Politics and the Theory of Organization. Journal of Law, Economics, and Organization 7, 106-129.

Moe TM, Wilson SA (1994) Presidents and the Politics of Structure. Law and Contemporary Problems 57(2), 1-44.

Obinger H, Schmitt C, Zohlnhöfer R (2014) Partisan Politics and Privatization in OECD Countries. Comparative Political Studies 47, 1294-1323.

Petrocik JR (1996) Issue Ownership in Presidential Elections, with a 1980 Case Study. American Journal of Political Science 40, 825-850.

Pollitt C, Bathgate K, Caulfield J, Smullen A, Talbot C (2001) Agency Fever? Analysis of an International Policy Fashion. Journal of Comparative Policy Analysis: Research and Practice 3, 271-290.

Potoski M (1999) Managing Uncertainty through Bureaucratic Design: Administrative Procedures and State Air Pollution Control Agencies. Journal of Public Administration Research and Theory 9, 623-640.

Rogoff K (1985) The Optimal Degree of Commitment to an Intermediate Monetary Target. The Quarterly Journal of Economics 100, 1169-1189.

Schoenfeld D (1982) Partial Residuals for the Proportional Hazards Regression Model. Biometrika 69, 239-241.

Shepsle KA (1991) Discretion, Institutions, and the Problem of Government Commitment. In: Bourdieu P, Coleman JS (eds) Social Theory for A Changing Society, pp. 245-265. Westview Press, Boulder.

Spiller PT (1993) Institutions and Regulatory Commitment in Utilities' Privatization. Industrial and Corporate Change 2, 387-450.

Steenbergen MR, Marks G (2007) Evaluating Expert Judgements. European Journal of Political Research 46, 347-366.

Tavares J (2004) Does Right or Left Matter? Cabinets, Credibility and Fiscal Adjustments. Journal of Public Economics 88, 2447-2468.

Thatcher M (2002) Delegation to Independent Regulatory Agencies: Pressures, Functions and Contextual Mediation. West European Politics 25(1), 125-147.

Tsebelis G (1995) Decision Making in Political Systems: Veto Players in Presidentialism, Parliamentarism, Multicameralism and Multipartyism. British Journal of Political Science 25, 289-325.

Tsebelis G (2002) Veto Players: How Political Institutions Work. Russell Sage Foundation/ Princeton University Press, New York/Princeton.

van der Brug W (2004) Issue Ownership and Party Choice. Electoral Studies 23, 209-233.

Van Thiel S (2012) Comparing Agencies Across Countries. In: Verhoest K, Van Thiel S, Bouckaert G, L®greid P (eds) Government Agencies: Practices and Lessons from 30 Countries, pp. 18-26. Palgrave Macmillan, Basingstoke.

Verhoest K, Van Thiel S, Bouckaert G, L®greid P (eds) (2012) Government Agencies: Practices and Lessons from 30 Countries. Palgrave Macmillan, Basingstoke.

Vogel SK (1996) Freer Markets, More Rules. Regulatory Reform in Advanced Industrial Countries. Cornell University Press, Ithaca.

Walgrave S, Lefevere J, Tresch A (2012) The Associative Dimension of Issue Ownership. Public Opinion Quarterly 76, 771-782.

Wonka A, Rittberger B (2010) Credibility, Complexity and Uncertainty: Explaining the Institutional Independence of 29 EU Agencies. West European Politics 33, 730-752.

Zohlnhöfer R, Obinger H, Wolf F (2008) Partisan Politics, Globalization, and the Determinants of Privatization Proceeds in Advanced Democracies (1990-2000). Governance 21, 95-121.

Appendix

Table A1 Determinants of agency establishment (models with country-fixed effects)

Model 5 Model 6 Model 7 Model 8

Cabinet ideology 1.083 1.049 1.162

(1.01) (0.62) (1.60)

Cabinet extremeness 1.398* 1.428** 1.414**

(1.92) (2.07) (1.97)

Cabinet ideology X social regulation 0.744** (-2.19)

Social regulation 1.720 (0.64)

Political constraints 0.863** 0.875** 0.872** 0.887*

(-2.47) (-2.20) (-2.23) (-1.90)

Political constraints X ln (time) 1.054** 1.048** 1.049** 1.042*

(2.56) (2.24) (2.28) (1.92)

Average cabinet duration 1.527 1.090 1.010 1.016

(0.51) (0.10) (0.01) (0.02)

Average cabinet duration X ln (time) 0.800 0.929 0.950 0.954

(-0.73) (-0.23) (-0.16) (-0.15)

Liberalization 1.996 2.043 2.071 1.728

(1.28) (1.32) (1.36) (1.00)

Privatization 1.520 1.360 1.294 1.584

(0.65) (0.46) (0.38) (0.71)

Diffusion (logged, by sector) 0.000* 0.000* 0.000* 1.322

(-1.92) (-1.90) (-1.90) (0.21)

Diffusion (logged, by sector) X ln (time) 129.815** 123.450** 121.921**

(2.51) (2.49) (2.49)

EU directive 92/44 1.165 1.275 1.295 1.050

(0.19) (0.29) (0.31) (0.06)

EU directive 97/51 13.069*** 11.378*** 10.854*** 13.953**

(2.83) (2.64) (2.59) (2.86)

EU directive 96/92 0.578 0.587 0.590 0.709

(-1.07) (-1.04) (-1.04) (-0.68)

2220 2220 2220 2220

N (country-sectors) 123 123 123 123

N (agencies created) 110 110 110 110

Log likelihood -405.2 -403.3 -403.5 -399.9

Note: Figures are hazard ratios from Cox t-statistics in parentheses; *P < 0.1, **P < 0,

proportional hazard regressions with country-fixed effects; 05, ***P < 0.01. EU, European Union.

0.10-1

® -0.05-1

-0.20 J

10 15 20 Survival time (years)

Figure A1 Joint effect of political constraints and political constraints X ln (time).

Note: Based on Model 3. Dotted lines represent 90-percent confidence intervals. For calculation of standard

errors, see Golub and Steunenberg (2007).

10 —| 50-5-10-15-20-

10 15 20

Survival time (years)

Figure A2 Joint effect of diffusion and diffusion X ln (time).

Note: Based on Model 2. Dotted lines represent 90-percent confidence intervals. For calculation of standard errors, see Golub and Steunenberg (2007).

c 0 -0.5

0 -1.0-

(1) 0 -1.5-

0.250.20-.S 0.15-

5 0.10-

0.05 0.00

Cabinet ideology

I-1-1-1-1-1-1—

AT BE CH CY DE DK ES

~~I-1—

~~1-1-1-1-1-1-1-1

IT LU MT NL NO PT SE UK

Omitted country

Cabinet ideology (squared)

I-1-1-1-1-1-1—

AT BE CH CY DE DK ES

—I-1—

—i-1-1-1-1-1-1-1

IT LU MT NL NO PT SE UK

Omitted country

0.0 -0.1 ■ i -0.2-ë -0.3° -O.4. -0.5 -0.6-

Cabinet ideology x social regulation

1-1-1-1-1-1-1—

AT BE CH CY DE DK ES

—1-1—

—1-1-1-1-1-1-1-1

IT LU MT NL NO PT SE UK

Omitted country

Figure A3 Coefficients from jackknife analysis.

Note: Graphs show coefficients and 90-percent confidence intervals from jackknife analysis, omitting one country from the sample at a time. Grey areas and lines represent coefficients and 90-percent confidence intervals from full models. Calculations based on Model 4.