Independent agencies? Political vulnerability and affinity of their leaders

Abstract

Are the leaders of independent agencies independent in practice? Are the independence requirements set out in legislation a guarantee of de facto independence? This paper reveals the relationship between de iure independence and de facto independence of independent agencies through two dimensions: political affinity and political vulnerability of their leaders. Our analysis reveals how the de iure independence of an agency affects the probability that agency heads will have connections to political parties and whether their mandates will end prematurely in a period of political transition, i.e., when a new government takes office. It also determines whether the biographical profile of agency heads (PhD degree, bureaucratic background, and political affiliation) can influence their security of tenure when governments change, and hence their independence. This is supported by an empirical evaluation of independent authorities in the Czech Republic between 1993 and 2021.

Introduction

Since the 1970s, institutions endowed with independence, i.e., agencies to which various powers have been delegated and which operate at arm’s length from elected politicians or government (Majone 1994, 1997), have diffused across the world (Levi-Faur 2005; Jordana et al. 2011; Gilardi 2005a, 2008).

Formally independent bodies were first set up in the field of finance at the level of central banks. Subsequently, so-called IRAs (independent regulatory agencies) emerged as typical regulators in network industries due to market-liberalizing processes such as deregulation and privatization. Over the years, they have expanded across sectors into public and social services. Competition, data protection, nuclear safety, food safety, or environmental protection authorities, among others, are often established as independent agencies (IA). The proliferation of IAs has led to many scientific investigations. Independence as an immanent feature of these institutions has been examined by abundant research. In this context, the literature most often deals with three main issues: the theoretical justification for the establishment and spread of IAs, aspects of formal (de iure) independence, and the relationship between formal (de iure) and actual (de facto) independence. The third area remains the least explored, and there is a lack of research that uses more recent data on IAs. This gap invites further research. Indeed, the abundant theoretical background on this issue opens unique opportunities for further investigation with up-to-date empirical data.

This paper analyzes whether de iure independence also leads to a greater degree of de facto independence and what other factors may influence de facto independence of IA heads. Except for small nuances, scientists find consensus (Guardiancich and Guidi 2016) on what constitutes the essence and features of de iure independence. The form and extent of de iure independence are always the result of rules of law. While de iure independence stems from clearly defined legal factors, de facto independence reflects the extent to which the institution is independent in practice. Assessing and conceptualizing de facto independence is challenging and can take many forms. As a contribution to this debate, this paper emphasizes another point of view, which is often overlooked. Rather than assessing real independence through policy outcomes, we argue that it may be insightful to look at de facto independence of IA heads in exercising their power. Ultimately, the heads of these agencies are behind the final policy decisions. Rules on the independence of an IA head or board members are a key part of the assessment of formal independence. The law often sets out rules on an appointment, term of office, restrictions on reappointment, and the prohibition on being a member of a political party. In light of these rules, it seems apt to focus on examining how individual heads are independent in practice. Subsequently, it will be possible to find a link between their de facto independence and the formal independence of the agency. This study considers not only individual CEOs but also board members if a collegial body heads the IA.

This research assesses formal independence using the Gilardi index (Gilardi 2002) compiled for each agency. It takes into account all changes in legislation in the monitored years. De facto independence is assessed through political party affiliation and the political vulnerability of IA leaders. Political vulnerability represents how IA heads survive changes of governments. Political affiliation reflects whether the IA head is linked to a political party. The aim of the de iure independence rules is to isolate certain public sector institutions from political interference, thereby ensuring the independence of IA leadership. Thus, it makes sense that IA heads should not be affiliated with politicians through political parties. Further, these heads should not be subjected to pressure to leave their positions when a new government comes in.

Our research explores de facto independence based on two parameters: political party affiliation and political vulnerability. Specifically, it forms five research hypotheses, which are examined by two logistic regression models.

H1

Stronger de iure independence increases the probability of no political ties.

H2

Stronger de iure independence decreases the probability of a prematurely finished mandate.

H3

The political affiliation of an agency head increases the probability of a prematurely finished mandate.

H4

PhD degree of an agency head decreases the probability of a prematurely finished mandate.

H5

The bureaucratic background of an agency head influences the probability of a prematurely finished mandate.

We use an original dataset in this study. It includes 10 independent authorities in the Czech Republic that are monitored from 1993 to 2021 (since the establishment of the Czech Republic as an independent state to the present). The Czech Republic serves as an example of a country that has transitioned from a totalitarian society to a democracy and from a centrally planned economy to a market economy. The Czech Republic, but also other post-communist countries, has been neglected in previous research on IAs. Our research seeks to fill this gap.

The paper is structured as follows. “Milestones in measuring independence” section reviews the literature on different concepts and measures of IA independence. “Independent agencies: evaluating political vulnerability and party affiliation of their heads” section forms and discusses five hypotheses for explaining the link between de iure and de facto independence. “Independent authorities in the Czech Republic” section describes and analyzes the data on independent agencies in the Czech Republic and their mandates. The discussion is embodied in “Discussion” section, followed by conclusions in “Conclusion” section.

Milestones in measuring independence

With regard to the issue of de iure independence, the work of Fabricio Gilardi is crucial (Gilardi 2002, 2005a, b, 2008). Gilardi’s concept of formal independence refers to the provisions set out in laws, by-laws, and agency statutes, which aim to guarantee the agency’s independence from elected politicians and government members. Gilardi is the author of an index developed to measure the formal independence of IRAs. The degree of formal independence reflects various aspects of independence that are enshrined in the legal instruments. Creating an independence index was an essential step for any comparative research of IRAs, because their institutional design is characterized by “extreme empirical heterogeneity” (Gilardi 2002, p. 874). Gilardi constructed a dataset with information on regulators from five sectors in seven European countries and measured their independence using his newly created index. Prior to Gilardi’s index, procedures for assessing independence had only been developed for the special case of central banks (Cukierman 1992; Cukierman et al. 1992; Cukierman and Webb 1995). The Gilardi index is based on them, but it is adapted to be applied to all independent agencies in general, regardless of the sector in which they operate.

The index consists of five main sections—the agency head status, the management board members’ status, the general framework of relationships with the government and the parliament, financial and organizational autonomy, and the extent of regulatory competencies. The vast majority of studies use this index or at least build on it, and some also use Gilardi’s original dataset (e.g., Maggetti 2007; Hanretty and Koop 2013; Koop and Hanretty 2018; Wonka and Rittberger 2010; Guardiancich and Guidi 2016; Ennser-Jedenastik 2016). Although the Gilardi index is widely used to measure de iure independence, there are some criticisms of its composition (e.g., in Hanretty and Koop 2012, 2013; Wassum and De Francesco 2020).

The consequences of the institutional setting, in other words, the degree of formal independence, can also be examined through the concept of de facto independence. However, the assessment of the IRA’s de facto independence faces considerable challenges. The first step is to develop certain comparative indicators, which is a highly demanding task. Thus, only a few scholars have found an empirical link between de iure and de facto independence. Maggetti (2007) and Hanretty and Koop (2013) can be considered the most fundamental studies, but both of these analyses arrive at conflicting findings.

According to Maggetti (2007), de facto independence is represented by the effective independence of agencies in their day-to-day regulatory actions. This conception is based on Majone’s definition of independent regulatory agencies as “highly specialized organizations enjoying considerable autonomy in decision-making” (Majone 1997, p. 152). In his research, Maggetti (2007) also focused on measurements of de facto independence from politicians. The nature of the relationship with the politicians is captured by six indicators of self-determination—frequency of revolving door; the frequency of ad hoc contacts such as internships, collaborations, and regular meetings; influence over the budget; influence over the internal organization; the weight of partisan membership on board members’ nominations; and political vulnerability (earlier departures of board members)—and by one indicator of the autonomy of regulation (active participation in the rule-making process). The indicators are then incorporated into the survey questions for the investigated authorities. The conclusion of Maggetti’s study is that “high formal independence is neither a necessary nor a single sufficient condition for high de facto independence from politicians” (Maggetti 2007, p. 279) and that some agencies may be strongly de facto independent, regardless of their formal independence.

On the other hand, Hanretty and Koop (2013) demonstrate that formal independence significantly predicts actual independence. They focus exclusively on independence from politics. Their study uses two indicators of de facto independence: a measure of political vulnerability (VUL) and a measure of the average turnover of the agency’s chief executive (TOR). The political vulnerability proxy, VUL, represents the percentage of government changes followed within 6 months by a change in the agency’s chief executive. The turnover proxy, TOR, is the reciprocal of the average tenure, in years, of chief executives of an agency. The findings of their analysis of 87 IRAs demonstrate that the de facto independence of IRAs can be explained reasonably well by the degree of their de iure independence, but other variables, such as the rule of law and the number of veto players, matter as well.

In another study, Fernández-i-Marín et al. (2016) adapted VUL and TOR to assess de facto independence of board members of regulatory agencies in Spain. De facto independence is thus also assessed through the political vulnerability of agency chiefs (whether they remain in the agency when governments change). They distinguish formal and de facto independence by combining the legally expected duration of an appointment and its real length. Specifically, they focused on whether certain personnel characteristics or professional biography of regulators affect their continuance in the agency. The study yielded unexpected results. Contrary to expectations, it turned out that regulators with political links are less vulnerable to political cycles. Professionals and experts as regulators, on the other hand, turned out to be more politically vulnerable; their knowledge and expertise do not give them protection in times of political transition.

Another important milestone for this paper is Thatcher (2005). Although he does not explicitly work with de facto and formal independence concepts, the connection to them in his analysis cannot be overlooked. In light of the principal-agent theory, he assesses the implications of using formal controls by elected politicians on regulators. The power of elected politicians to appoint board members is an essential means of control, but it can also be a tool to put politically connected people in a regulatory agency. His analysis of data on regulators from 4 European countries concludes that elected politicians make limited use of their powers of appointment to put affiliated people in the agency. They also do not push for early termination of regulators’ mandates, allowing them to remain in the agency until the end of their term. Statistically, in his study, the proportion of people with political connections in regulatory bodies varied from country to country. In comparison, the proportion of those who terminated their mandate prematurely (political vulnerability) was relatively low in most cases.

Finally, Ennser-Jedenastik (2016) put forward the first large-N study about the relationship between de iure agency independence and party politicization. Surprisingly, the study shows that party politicization increases markedly with de iure independence. His analysis confirms the assumption “that higher levels of de iure independence create incentives for party politicization, thus potentially lowering de facto independence” (p. 515). Ennser-Jedenastik (2016) reveals that the expected level of attempts to put political affiliates into the agency is greater in highly de iure independent agencies. It is also greater when the regulator is the agency CEO than when the regulator is just a board member.

Independent agencies: evaluating political vulnerability and party affiliation of their heads

This study focuses on the independence from politicians. We deliberately leave aside independence from regulated entities and other factors. Our research explores de facto independence based on two parameters: political party affiliation and political vulnerability.

Both the literature and the legislation suggest attempts to limit the links of agency holders to political parties and to limit pressures for frequent personnel changes in the IA leadership. These efforts are intended to lead to higher de facto independence of the agency. For example, in many cases, the legislation explicitly prohibits IA officials from being members of political parties, introduces fixed terms of office, or limits possibilities for removal. Therefore, it is desirable that IA heads do not have political ties and are not prematurely removed from their positions. In this research, we also assume that low political vulnerability and leaders with no political affiliation will represent higher de facto independence of the agency.

Hypothesis 1—De iure independence and political affiliation: stronger de iure independence increases the probability of no political ties

It can be assumed that it is in the interest of government that an agency endowed with a high degree of formal independence should also operate independently in practice. This argument stems from the theoretical justification for the creation of independent authorities. Its cornerstone is that the government will solve the problem of time-inconsistent preferences (Kydland and Prescott 1977) and the problem of making a credible commitment by delegating some of its powers to an independent authority. It originally appeared at the central banking level (Barro and Gordon 1983; Rogoff 1985) and only later began to be applied to independent authorities.¹

The independent regulator or IA head that is insulated from short-term politics and political cycles will be better able to meet its long-term policy commitments. Regulatory objectives or policy strategies can be set for a longer period than the electoral term and will not be distorted by the pursuit of short-term electoral gains. The regulator’s desire for re-election will not influence the regulatory policy. Based on this theory, it is not advantageous for the government to put persons with political ties in IA leadership but rather apolitical experts.

For simplicity, it can be argued that if there are no people with political affiliation in the leadership of the independent authorities, the authority can be considered de facto independent. People with no political affiliation and no ties to politics will not be subordinated to party goals and political pressures. These connections may not be exclusively linked to the ruling political party. Under a proportional electoral system, it is common that the distribution of leadership positions is subject to negotiation and may reflect the party plurality resulting from the elections. We will therefore examine ties and affiliations to any political party.

In addition, there are certain tools that, by their very nature, may also reduce the tendency to appoint politically connected persons to independent agencies. Legislation that creates a high degree of formal independence will inherently regulate the political connections of board members. For many independent authorities, the government is not the body that appoints the head of the agency or the board members. Instead, the appointing authority may be, for example, the president, parliament, or an NGO. In some cases, the authority may choose its own head. It is therefore more difficult for the government to put related people in the agency. Another parameter of formal independence is the explicit prohibition for board members to hold a position in a political party.

A high degree of formal independence, i.e., the exercise of powers by an independent authority, may be more advantageous to the government (because of its greater credibility), even at the cost of losing some of the delegated powers. The government can also use the IA as a tool to shift blame, for example, when problems appear. Therefore, efforts to capture seats in the independent authority are not worthwhile for the government. On the other hand, the reverse tendency may also occur. A high degree of formal independence of the agency may be a signal for the government to start attempting to control the agency. As formal independence increases, it will be more difficult for the government to put its own people in place in future. If formal independence provides a stable and long-term position for the members of the independent authority, it could be disadvantageous for politicians to lose their powers. In his study, Ennser-Jedenastik (2016) also assumed that higher levels of de iure independence generate incentives to appoint ideologically like-minded individuals to the agency. Higher de iure independence would thus mean greater party politicization. The results of empirical research will show which tendencies prevail.

Hypothesis 2—De iure independence and political vulnerability: stronger de iure independence decreases the probability of a prematurely finished mandate

The government’s attempts to install related persons in independent agencies go hand in hand with whether the incumbents will be able to complete their mandate to the end properly. The terms of office of heads and board members in independent agencies are usually fixed. Fixed terms are intended to provide the IA head with greater independence and a greater degree of certainty. The IA head will be able to plan a long-term strategy and achieve more elaborate goals in a predetermined period than in an indefinite period. Therefore, formal independence is rated higher when the fixed term is longer.

The length of tenure is often designed not to coincide with changes of government so that the incoming head will more likely not be an expression of the new government (Hanretty 2010). The term of office is usually deliberately different from the length of the political cycle of the government so as to reduce the effort to replace people when a new government takes office. The legislation also often states that the government (or other appointing authority) cannot dismiss the heads of independent agencies during their term of office. Alternatively, the government may remove the head, but only under certain strict conditions and only for reasons unrelated to policy. However, all these provisions are not sufficient guarantees leading to the proper completion of the mandate. When a head leaves prematurely, there is no way to prove that the head resigns voluntarily or under hidden pressure from politicians. Forcing the head to resign may represent some new constraint connected to the government. Leaving on his or her own accord may reflect a belief that the government should have a ‘clean slate’ to influence the forthcoming selection of a chief executive (Hanretty 2010). In many cases, the motivation to leave a position stems from political circumstances. We, therefore, assume that data on whether board members and heads do not end their mandates prematurely after the political change can serve as a suitable indicator of de facto independence. The basic idea is that when tenure rules are not followed, and there are premature departures from the agency when a new government takes office, the heads have no opportunity to behave independently in their practice. In addition, Fernández-i-Marín et al. (2016) show that the professional characteristics of board members can play an important role in how tenure rules are observed in practice.

Hypothesis 3—Political affiliation and political vulnerability: the political affiliation of an agency head increases the probability of a prematurely finished mandate

We assume that IA heads with political ties will align themselves with the objectives of their political masters and thus become less independent. Then, when their masters change (a new government takes office), the political vulnerability of incumbent heads will be higher. The new political masters will want to staff the agencies with their kindred servants while heads with no political ties remain in the agency. Thus, to ensure greater stability and independence and mitigate the turnover of agency heads, it would be preferable for the heads to have no political affiliations.

Hypothesis 4—Expert status and political vulnerability: PhD degree of an agency head decreases the probability of a prematurely finished mandate

The general perception is different about heads with expert status (persons with a scientific or academic background). Rather than adapting it to political pressures, they are expected to apply regulatory policy based on their expertise, in a stable and predictable manner. This is also supported by the traditional analysis of the delegation problem. Politicians are aware that they have neither sufficient expertise nor the capacity to adapt the policy to ever-changing conditions and circumstances, and thus, they delegate some of their powers to specialized bodies led by neutral experts (Majone 1996). Hence, the new government will make little effort to replace them. Their political vulnerability will therefore be lower and their de facto independence stronger.

Hypothesis 5—Bureaucratic background and political vulnerability: the bureaucratic background of an agency head influences the probability of a prematurely finished mandate

Lastly, the situation will be interesting for persons whose life biography is linked to the position of a bureaucrat. On the one hand, persons with a bureaucratic background may appear as neutral as the experts mentioned above since they are not tied to any political party and their function is professional. On the other hand, they may tend to reduce their political vulnerability (trying to keep their position) and thus may adapt their behavior to the preferences of the new political masters. In the first case, the low political vulnerability of bureaucrats would be a desirable phenomenon. Proper mandate completion, free from political pressures, will ensure that bureaucrats are apolitical and professional in their policy-making and thus have greater de facto independence. However, in the latter case, the low political vulnerability could be evaluated negatively. To defend their current position, bureaucrats may attempt to accommodate policy to the new masters. After a political change, bureaucrats then may want to reduce their political vulnerability by conforming to the new government’s goals. They seek less political vulnerability and adjust their preferences to incoming politicians rather than lose their jobs. In such cases, heads with bureaucratic backgrounds may weaken de facto agency independence (Fernández-i-Marín et al. 2016). Theory gives ambiguous predictions and resolving them is an empirical matter.

Independent authorities in the Czech Republic

In this study, we examine10 independent authorities in the Czech Republic for the period 1993–2021 (see Table 1). We choose 1993 because the Czech Republic was established as an independent state on 1 January 1993. With the transition to a market economy, new agencies began to emerge. It was necessary to cover agendas that had no place in a centrally planned economy (free TV broadcasting, competition protection, etc.). As a result of privatization and liberalization of public utilities, independent regulators of network industries emerged, with the increase in their independence supported by EU legislation² after the Czech Republic became an EU Member State. The number of independent agencies has continued to increase in recent years. In this paper, we draw on the definition from Thatcher and Stone Sweet (2002, p. 2). First, all the authorities surveyed are endowed with their own powers and responsibilities by public law. Second, they are organizationally separate from the ministries. Third, their leaders are neither directly elected nor directly controlled by elected officials. The authorities we have selected are either explicitly designated by law as independent or the law states that the authority is guided in its activities only by the law and not by instructions from the government or other institutions.

Table 1 Independent agencies in the Czech Republic

The size of the agency, indicated by the number of staff, represents, in other words, the resources of the agency. The more resources an agency has, the more interesting it will be to politicians. We expect that the government would want to control policy in large agencies rather than smaller ones. On the other hand, it may be more difficult to influence a larger institution because of its more robust and complex organization.

We determined the Gilardi index (Gilardi 2002) for each independent authority every year from their foundation. The Gilardi index determines the numerical value of formal independence (for the purposes of Hypotheses 1 and 2). Despite criticism of this index, we use it in its original form. This ensures that the observed extent of formal independence is comparable to other research. To this end, we analyze laws, government ordinances, agency statutes, and national legal repositories. We find that formal independence varies considerably among the authorities surveyed.

Our dataset contains information about heads of agencies or board members (when the agency is headed by a collegiate body). We assume political affiliation and vulnerability to be higher for individuals who are CEOs than for board members.

The first step in the research was to identify all agency holders, their start date and end date (assumed and actual). In terms of personal background, we investigated whether representatives are publicly affiliated with a political party. Our evaluation was based on similar criteria as chosen by Ennser-Jedenastik (2016). Political affiliation was assessed according to the following criteria: (1) membership in a political party, (2) holding public agency for a political party, (3) membership in a political party auxiliary group, and (4) publicly speaking in favor of a political party. In practice, most cases fulfilled the first criterion; the remaining cases were rare.

For agency heads who had already completed their mandate with a fixed term of office, we examined whether their mandate ended properly or whether it ended prematurely during a period of political transition. We define the period of political transition as the period of one year after the new government’s appointment (i.e., the day the new government takes power).

In terms of other biographical data, we ascertained whether heads had a bureaucratic background, i.e., whether they had worked as a civil servant for a significant part of their career. It was also examined whether they had obtained a PhD or higher degree.

The sources of this information included media archives, publicly available online sources and databases, ordinances of the government and other bodies, agency documents and statutes, and annual reports. Where information was missing, it was necessary to contact the agency directly to request information.

Two datasets were collected. The first dataset serves for the analysis of the agency’s de iure independence and political affiliation of IA heads (H1), while the second dataset serves for the analysis of political vulnerability of IA heads (H2, H3, H4, H5). Because the second dataset addresses whether or not the mandate ended prematurely, we exclude IA heads who do not have a fixed term of office as well as IA heads who are currently in the agency.

Data description for H1: political affiliation

The first dataset contains 181 observations, each corresponding to one mandate, and 4 variables. The dependent variable $Political\_affiliation$ is a binary variable where 1 indicates representatives with political affiliation and 0 otherwise. The key explanatory variable $Independence\_agency$ is a value of Gilardi’s index ranging from 0 to 1, with 1 indicating that the independent authority is fully independent and 0 indicating the highest dependence. The other explanatory variables are $Size$, which stands for the number of employees of the agency, and $CEO$, which is a binary indicator where 1 stands for the CEO and 0 for the board member. All variables are measured at the beginning of a mandate.

Table 2 shows the numbers of observations for binary variables $Political\_affiliation$ and $CEO$. Figure 1 shows histograms for continuous variables $Independence\_agency$ and $Size$. It also depicts $\log \left( {Size} \right)$, which stands for the logarithmic transformation of $Size$ and is further used in the model since the original variable $Size$ is not suitable for the modeling as it exhibits right-skewed distribution due to a small number of very large agencies. The dataset consists of information about 144 independent mandates and 10 agencies, which form individual clusters in further analysis.

Table 2 Number of observations for binary variables

Fig. 1

Histograms for the continuous variables of agency independence (Independence_agency), size, and the logarithmic transformation of size (log(Size))

Data description for H2, H3, H4, H5: political vulnerability

The second dataset contains 135 observations, each corresponding to one mandate, and 6 variables. The dependent variable $Mandate$ is a binary variable where 1 indicates the prematurely finished mandate (in the period of political transition) and 0 indicates the completed mandate. The explanatory variables $dependence\_agency$, $Political\_affiliation$, $Size$, and $CEO$ are defined in the same manner as in dataset 1. The newly introduced binary variables $PhD$ and $Bureaucrat$ indicate whether the representative has a PhD degree and a bureaucratic background. All variables are measured at the end of a mandate.

Table 3 shows the numbers of observations for binary variables $Mandate$, $Political\_affiliation$, $CEO$, $PhD$, and $Bureaucrat$. Three representatives have missing values for the $Bureaucrat$ variable because the relevant data could not be found. Of those not currently in the agency, 16% ended their mandate prematurely, in the period of political transition. CEOs account for 30% and board members 70% of the total number of appointees. Agency heads most often have bureaucratic backgrounds; 30% come to the independent agency from other government agencies or ministries. Holders of a PhD or higher academic degree account for 27%. Finally, 36% of the representatives were publicly connected with a political party. Ennser-Jedenastik (2015) arrives at an almost identical finding, namely that 34% of 303 heads of 100 West European regulatory agencies had a political party affiliation. Thatcher (2005) states that political affiliation ranged from 3% in Britain, 36% in Germany, and 46% in France to 77% in Italy.

Table 3 Number of observations for binary variables

Figure 2 shows histograms for continuous variables $Independence\_agency$ and $Size$. It also depicts $\log \left( {Size} \right)$, which stands for the logarithmic transformation of $Size$ and is further used in the model since the original variable $Size$ is not suitable for the modeling as it exhibits right-skewed distribution due to a small number of very large agencies. The dataset consists of information on 104 independent mandates and 7 agencies, which form individual clusters in further analysis.

Fig. 2

Histogram for the continuous variables of agency independence (Independence_agency), size, and the logarithmic transformation of Size (log(Size))

Analysis H1: De iure independence and party affiliation

To test Hypothesis 1, we specified a logistic regression model. The dependent variable is a binary indicator of political affiliation, i.e., whether the representative is publicly affiliated with a political party or not. The independence of the agency, the size of the agency, and a binary variable that indicates whether the person is the agency’s CEO or just a member of the board serves as explanatory variables. Ennser-Jedenastik (2016) works with political constraints and the rule of law. By focusing on independent agencies within a single country, we consider these variables sufficient for the purpose of the paper.

Formally, the conditional probability of being publicly affiliated with a political party is defined as

$$p = P(Political\_affiliation = 1 | independence\_agency, \log \left( {Size} \right),CEO).$$

Then our proposed logistic model is

$$logit\left( p \right) = \beta_{0} + \beta_{1} Independence\_agency + \beta_{2} \log \left( {Size} \right) + \beta_{3} CEO,$$

where $\beta_{0} , \beta_{1} , \beta_{2}$ and $\beta_{3}$ are parameters and $logit\left( p \right) = \ln \left( {p/\left( {1 - p} \right)} \right)$, i.e., log-odds. The parameters are estimated by the maximum likelihood method. In “Appendix,” we present alternative specifications of the model.

Table 4 shows the results for the defined logistic model and its special case with the restriction $\beta_{3} = 0$. Standard errors are estimated by the clustered sandwich estimator. Clustered sandwich estimators are used to adjust inference when errors are correlated within (but not between) clusters. We used the approach of Cameron et al. (2011), which allows for clustering in multiple cluster dimensions. Table 4 includes the clustered standard errors, where the unique identifier for individual agencies and independent mandates form clusters.

Table 4 Predicting the probability of political affiliation of IA heads

The positive effect of the control variable $CEO$ is in line with the expectation, but it is not statistically significant using any reasonable significance level ($p$ value = 0.7282). Moreover, the model without $CEO$ is preferred over the full model based on the Akaike information criterion. Thus, we exclude the variable from the regression and interpret the restricted model. The coefficient for $Independence\_agency$ is significantly negative; thus, a higher degree of independence of the agency decreases the probability of being politically affiliated. Similarly, since the coefficient for $\log \left( {Size} \right)$ is significantly negative, representatives of the larger agencies have a lower probability of being politically affiliated than the representatives of the small agencies.

Figure 3 shows the probabilities of political affiliation for different values of Gilardi’s index and sizes of the agencies—the smallest (4 employees), average (447 employees) and largest (2,464 employees) agency in our dataset.

Fig. 3

Estimated probabilities of political affiliation for different values of Gilardi’s index and agency size

Analysis H2 H3 H4 H5: De iure independence, biographical profiles, and political vulnerability

For Hypotheses 2 to 5, the dependent variable is a binary indicator of political vulnerability where 1 indicates that the representative was prematurely dismissed or left the agency in a period of political transition and 0 denotes that the representative has completed his or her mandate. The explanatory variables are as follows: the independence of the agency; the political affiliation of the representative; the size of the agency; and binary variables that indicate whether the person is the agency’s CEO, has a PhD degree, and has a bureaucratic background (see “Data description for H2, H3, H4, H5: political vulnerability” section for further information).

Formally, the conditional probability of a prematurely finished mandate is defined as

$$p = P(Mandate = 1 | x),$$

where $x$ are explanatory variables. Then our proposed logistic model is

$$logit\left( p \right) = \beta_{0} + \beta_{1} Independence\_agency + \beta_{2} Political\_affiliation + \beta_{3} \log \left( {Size} \right) + \beta_{4} CEO + \beta_{5} PhD + \beta_{6} Bureaucrat,$$

where $\beta_{0}$–$\beta_{6}$ are parameters and $logit\left( p \right) = \ln \left( {p/\left( {1 - p} \right)} \right)$, i.e., log-odds. The parameters are estimated by the maximum likelihood method. In “Appendix,” we present alternative specifications of the model.

Table 5 shows the estimated coefficients for the defined logistic model and its special case with the restrictions $\beta_{4} = \beta_{5} = \beta_{6} = 0$. Table 5 also includes the standard errors estimated by the clustered sandwich estimator, where the unique identifier for individual agencies and independent mandates form the clusters.

Table 5 Predicting the probability of prematurely finished mandate of IA heads

The positive effect of the control variable $CEO$ is in line with the expectation, but it is not statistically significant using any reasonable significance level ($p$ value = 0.3697). In the same manner, the negative effect of the variable $PhD$ and positive effect of the variable $Bureaucrat$ are in line with the hypotheses, but the effects are not statistically significant using any reasonable significance level ($p$ values are 0.6401 and 0.6689, respectively).

The coefficients for $Independence\_agency$ and $Political\_affiliation$ are significantly negative suggesting that stronger independence of the agency and political affiliation of the representative decrease the probability of a prematurely finished mandate. Moreover, since the coefficient for $\log \left( {Size} \right)$ is also significantly negative, representatives of the larger agencies have a lower probability of prematurely finishing their mandates than the representatives of the small agencies.

Figures 4 and 5 show the probabilities of a prematurely finished mandate for different values of Gilardi’s index and sizes of the agency—the smallest (26 employees), average (394 employees), and largest (1698 employees) agency in our second dataset. Figure 4 shows the probabilities for representatives without political affiliation and Fig. 5 shows the probabilities for representatives with political affiliation.

Fig. 4

Estimated probabilities of a prematurely finished mandate for representatives without political affiliation for different values of Gilardi’s index and agency size

Fig. 5

Estimated probabilities of prematurely finished mandates for representatives with political affiliation for different values of Gilardi’s index and agency size

Discussion

H1 De iure independence and party affiliation

Our results confirmed the first research hypothesis. The more de iure independent an agency is, the less likely it is that a politically affiliated person will head the agency. Despite the general expectation that large agencies will attract politicians who will tend to put their political affiliates in those agencies, the opposite turned out to be true. Heads with political affiliations are less likely to be found in large agencies. Thus, it turns out that it is easier for politicians to control and install their people in small agencies than in large ones with complex and robust structures.

Moreover, politicians may assume that if they put their people in a small agency, the agency will indeed be under their control and in line with their interests and objectives. In a large agency, this may not always be the case, and an affiliated head may not be a definite guarantee of pursuing government objectives. Due to the more complex and extensive structures of a large agency, it might be more difficult even for an allied person to direct the policies of the agency exactly as the government wants. Small agencies are therefore more attractive to politicians.

H2 De iure independence and political vulnerability

The second research hypothesis was also confirmed; as the de iure independence of the authority increases, the probability that the head will not complete its mandate decreases. In addition, our results show that heads in large agencies have a lower probability of prematurely finishing their mandates than the representatives of the small agencies. For the heads of a large agency, the fact that they control large resources seems to be crucial. Benefits arising from large agencies are so great that the heads do not want to lose them. Thus, they would rather align the agency’s policies with the government’s preferences and objectives than lose these benefits. The incentives of the head of a large agency to reduce his political vulnerability by conforming to the demands of politicians will be greater than for the head of a small agency. In small agencies, the efforts to accommodate the government may not be as strong because the potential losses will not be as great as in a large agency.

H3 Political affiliation and political vulnerability

In this research hypothesis, we assume the general idea that persons with a political affiliation would be more politically vulnerable and thus provide the agency with less de facto independence. In line with common expectations, persons with political connections should be more prone to prematurely finishing their mandate because the incoming government will want to install its own affiliates. However, our results show the opposite. Heads with political affiliations are less likely to end their mandate prematurely after a new government takes office.

We then reach the same conclusion as Fernández-i-Marín et al. (2016), namely that “political links are actually a protection against removal after a political change occurs, while not having such ties increases the risks of being removed” (p. 241).

Thus, the position of politically affiliated individuals is more stable. This leads to ambiguous conclusions. In our research, low political vulnerability represents a higher level of de facto independence. However, it cannot be concluded that authorities will be more de facto independent if they are run by persons with political connections. Although politically affiliated heads are not dismissed prematurely, they may be under the influence of the government and subordinate policy to its demands.

H4 Expert status and political vulnerability

We anticipate that the specialized and expert knowledge of professionals with a scientific or academic background (those with a PhD or higher) will help these professionals to withstand political changes. Their expertise and apolitical neutrality will make them better able to defend their positions to politicians. If PhD holders were less politically vulnerable, they would be able to make regulatory policy in the long run without political pressures. The head of an agency who possesses a PhD would thus represent greater de facto independence. Our analysis shows that the fact that a head has a PhD reduces the probability of early termination, but the effect is not statistically significant. Thus, we did not confirm the fourth hypothesis.

H5 Bureaucratic background and political vulnerability

The positive effect of a bureaucratic background suggests that bureaucrats are more exposed to early departures, thus exhibiting greater political vulnerability. Contrary to common expectations, it suggests that bureaucrats cannot defend their positions once a new government takes office. This could also mean that bureaucrats do not avoid conflicts with politicians by accommodating the preferences of newly arrived political masters. However, as the effect is not statistically significant in our analysis, we cannot conclude that an agency head with a bureaucratic background will represent weaker or stronger de facto independence.

Conclusion

How are legal rules ensuring independence from politicians reflected in the actual practice of independent agencies?

First and foremost, if an agency is to be independent of politicians, it should not be led by a politically affiliated person. If the legal rules work in practice, the probability that an agency endowed with strong independence will be led by a politically affiliated head is low. This argument was fully confirmed in our research. Higher levels of de iure independence are associated with lower levels of political affiliation. In this regard, our findings refute the doubts about the effectiveness of independence rules as a tool to prevent political influence on independent agencies.

Furthermore, de iure independence rules ensure that IA heads will be able to carry out their functions uninterrupted until the end of their mandate so that they can independently pursue their long-term policy objectives. If these legal rules work in practice, the probability that the chief executive will be prematurely removed should decrease as the independence of the agency increases. Our research has again confirmed this.

Surprisingly, the analysis shows that politically affiliated heads are less at risk of leaving the agency when a new government takes office. This suggests that certain arrangements may persist between political parties that keep politically affiliated individuals in their seats and thus influence the personnel leadership of IAs. In the case of Czech independent authorities, political affiliation serves as a protection against premature dismissals.

All of our findings can serve as useful insights for the creation of legal rules concerning independent authorities and, for example, for the selection of their heads. The results of this study provide a unique picture of the effectiveness of formal independence rules and their functioning in the real life of independent authorities. Although we also tested the effect of CEO, expert and bureaucratic backgrounds, we found those effects to be insignificant. The inclusion of additional variables related to agency characteristics or other biographical profiles of leaders was beyond the scope of this paper but provides opportunities for further research.

Appendix: Robustness checks

We present alternative models for the robustness check, see Tables 6 and 7. First, we specify the models with a different definition of the agency size. Originally, we used the binary variable $SizeB$, where 1 means the small office with less than 150 employees and 0 otherwise. In the current model, we increased the accuracy of the information about the agency’s size and collected the new variable Size that measures the number of employees of the agencies. We also present a model without logarithmic transformation of the $Size$ variable.

Table 6 Predicting the probability of political affiliation of IA heads

Table 7 Predicting the probability of prematurely finished mandate of IA heads

Moreover, as pointed out by the reviewer during the review process, given the political turbulence in the 1990s, and to some extent again after the global financial crisis, it make sense to include decade dummies. Political appointments in the 1990s, when not only the agencies were new, but the state was, more generally, quite new, and heavily influenced by EU considerations and international reputation, are likely to have been subject to very different dynamics than appointments in the 2000s, and again in the 2010s. Thus, we also included and statistically tested dummy variables for the decades, $dec90s, dec00s, dec10s$, in order to capture these features.