2.1 Bureaucratic Autonomy

Finding an appropriate balance between accountability and autonomy is a major underlying question in public administration. On the one hand, holding civil servants, accountable for their actions is a central value in governance, especially democratic governance (Bovens, Goodin, and Schillemans Reference Bovens, Goodin and Schillemans2014; Finer Reference Finer1936). Ministers are supposed to hold their civil servants accountable, and then the legislature enforces accountability over the ministers. Even though accountability has become more complex and multifaceted (Olsen Reference Olsen2017) this more traditional form of accountability remains an important component of democratic governance.

On the other hand, however, the civil service is supposed to enjoy some autonomy and to be able to exercise independent judgment as it applies laws, and provides policy advice to the government (Lynn and Robichau Reference 60Lynn and Waters Robichau2013). The civil servants are expected to be experts in their tasks, and to employ that expertise to make the best possible decisions, and provide the “frank and fearless” advice that political leaders need. While civil servants are meant to serve their political leaders, they are also responsible for law and ethics, and are expected, again in the conventional description of democratic political systems, to act separately from political leaders (Friedrich Reference Friedrich, Friedrich, Beyer, Spero, Miller and Graham1935).

Beyond legal compliance, bureaucrats are also driven by their personal convictions and institutional loyalties. While Weber (Reference Weber1947) championed a formalistic impersonality – sine ira et studio (without hatred or passion) – it is well established that civil servants, like all individuals, hold personal preferences that inevitably shape their decision-making (Peters Reference Peters2018). These decisions may be influenced by self-interest, organizational priorities (Downs Reference Downs1965, Reference Downs1967), or even a commitment to safeguarding the state from undemocratic forces (Yesilkagit et al. Reference Yesilkagit, Bauer, Guy Peters and Pierre2024). Ultimately, the extent of bureaucratic autonomy and an individual’s willingness to exercise it determine how civil servants navigate their relationships with elected officials.

Although the autonomy of civil servants has been discussed most frequently concerning upper-level civil servants, it is also important for street-level bureaucrats who make decisions about clients, and who are engaged with local communities in adjusting policy (Nabatchi and Amsler Reference Nabatchi and Blomgren Amsler2014). Political leaders appear to be more concerned with whether senior civil servants will agree with them than with what actually happens when lower-level officials implement the programs.

The conflict between accountability and autonomy for civil servants has been central in public administration theory, and has been enshrined in the writings of Max Weber (Reference Weber, Roth and Wittich1978) and Woodrow Wilson (Reference Wilson1887), as well as numerous scholars coming after them. While both foundational scholars emphasized the importance of accountability, they also recognized the need for civil servants to be able to make their own decisions as they apply the law, and therefore to be free of political interference. The underlying logic was that, while implementing the law, civil servants should be able to make their own decisions and then be held accountable (Sager and Rosser Reference Sager and Rosser2009). Civil servants were assumed to be professionalized, with values and service commitments defined by their profession, and therefore could be trusted to make the right decisions in most instances.

This conflict of autonomy and accountability also appears in the familiar debate between Carl Friedrich (Reference Friedrich, Friedrich, Beyer, Spero, Miller and Graham1935) and Herman Finer (Reference Finer1936) over accountability, albeit phrased in terms of forms of accountability (see Heidelberg Reference Heidelberg2020). Friedrich (Reference Friedrich, Friedrich, Beyer, Spero, Miller and Graham1935) argued that the inner controls of civil servants, and their adherence to the law, were sufficient for accountability, and further that politicians were often poorly equipped to influence administrative decisions. This conception of the roles and capacities of civil servants would put them in a central position in governing, with political leaders legitimately having limited direct control over the decisions made by civil servants. This internalized version of accountability gives civil servants great autonomy. It can also put them in conflict with their nominal superiors in public organizations who can claim greater legitimacy because of their connection to the electoral process.

Herman Finer (Reference Finer1936), on the other hand, argued for more control from politicians, and more direct responsiveness of administrators to their political supervisors. In this view, the civil servant should do what he or she is told, and be willing to be over-ruled if they make a decision that does not conform to the wishes of the minister or other political leader. Thus, very little autonomy of action is acceptable, and the civil servant is directly accountable for actions. Consequently, in this view, the civil servant has little latitude for independent judgment about performing their tasks. It also implies some degree of politicization of the civil service, as they are expected to be highly responsive to their political superiors, rather than to their own professional values. When one encounters illiberal political leaders this view of accountability can also put them in the position of having to decide whether to accept illegal or immoral orders.

Phrased differently, Friedrich (Reference Friedrich, Friedrich, Beyer, Spero, Miller and Graham1935) was willing to accept more autonomy for civil servants, while Finer (Reference Finer1936) wanted more control and more direct intervention by politicians, or perhaps by the courts and other oversight bodies, to enforce accountability. Both components of this dichotomy of accountability and autonomy must be considered carefully as viable options for organizing government. On the one hand, it is important that policies are shaped by elected officials, and that the rest of government accepts the electoral mandate. On the other hand, however, civil servants are professionals and experts, so they should be empowered to make decisions within their sphere of competence.

To further explore bureaucratic autonomy, consider Bersch and Fukuyama’s (Reference Bersch and Fukuyama2023, 214) definition as

The definition includes at least five elements that merit discussion in the context of bureaucratic resistance. They are (1) the agents, that is, executive agencies, (2) discretionary authority, (3) ambiguity and incompleteness, (4) corruption, and (5) national interest.

Framing organizations as the agents of resistance implies unified thinking or, in other words, a consolidated organizational culture. Each agency will hold its specific interests and act to protect them. To illustrate, the U.S. Environmental Protection Agency (EPA) will likely use its powers to deter environmental crimes regardless of the presidential administration’s partisan leaning. Ozymy, Menard, and Jarrell (Reference Ozymy, Menard and Jarrell2021) find that the EPA has been capable of sustaining the stability of its criminal prosecutions over time despite the expectations of different presidents. This includes periods when the administration was clearly hostile toward the agency, for instance, those of Ronald Reagan and Donald Trump.

The Bersch and Fukuyama definition also makes the decisions made by civil servants to be unconstrained except for those five criteria, and to follow their “wishes.” While the delegation of authority to executive agencies does provide latitude, these decisions cannot be “arbitrary and capricious,” to use the language of the Administrative Procedures Act in the United States. Rather, those decisions must be in compliance with the public employee’s attributions, adequate technical standards, and the rule of law.

We posit that organizations are not the exclusive agents of decision-making. Despite organizational pressures, individuals hold their own values and may have significant discretion to affect policy implementation. Coyne and Hall (Reference Coyne and Hall2018), for instance, describe how public employees acted as policy entrepreneurs to push for relevant decisions impacting the current surveillance policy in the United States. In another study, Brazilian civil servants used their discretion, technical expertise, and political motivations to play a key role in the implementation of transportation policy and a river’s cleanup (Guedes-Neto Reference Guedes-Neto2022). Also, the street-level bureaucracy literature is filled with examples of individuals exercising their own discretion while implementing policy (Hupe Reference Hupe2015). In all examples, autonomous bureaucrats were not necessarily acting on behalf of their agencies, nor following the expectations of elected officials.

Discretionary authority, as mentioned before, creates a pathway for bureaucratic autonomy. Elected officials do not possess the time or knowledge to fully control the policy process, especially compared to the structural role and expertise of public employees (Downs Reference Downs1967; Huber and Shipan Reference Huber and Shipan2002). Politicians, therefore, delegate decisions to the bureaucracy, whose actions are likely to be influenced by a series of factors – politicians’ expectations being only one of them. After all, as Carpenter (Reference Carpenter2002) demonstrates, elected officials often know that going against the values of specific agencies may be inefficient and politically costly. One example regards the confrontations between the EPA and hostile presidential administrations. Prior research finds that bureaucrats relied on “hidden actions” within their legal possibilities to sustain their agenda (Ozymy, Menard, and Jarrell Reference Ozymy, Menard and Jarrell2021), for instance, transferring resources from one action to another (Ringquist Reference Ringquist1995) and using their statutory authority to set fines (Auer Reference Auer2008).

Incomplete or ambiguous policies may be easy targets for bureaucratic decision-making (Baier, March, and Saetren Reference Baier, March and Saetren1986). Again, it does not necessarily mean that they were poorly designed. It is possible that they are purposefully ambiguous to allow public employees to make decisions depending on the nature of times or, as Downs (Reference Downs1967) posits, based on their considerably higher expertise in such areas. These cases need not represent instances of bureaucratic resistance, since even Herman Finer would likely agree that some degree of autonomy is essential for policy implementation. Still, they facilitate resistance if bureaucrats decide to use them to go against the interests of elected officials.

Corruption and national interest are often both catalysts and focal points of bureaucratic resistance. Consider corruption first. When politicians exert significant control over public employees’ careers, they can more easily co-opt bureaucrats to facilitate rent extraction from the state (Brierley Reference Brierley2020). This dynamic helps explain why Dahlström, Lapuente, and Teorell (Reference Dahlstrom, Lapuente and Teorell2012) find that countries with stronger meritocratic recruitment systems tend to experience lower levels of corruption.

While bureaucratic autonomy can enable officials to misuse their discretion for personal gain (Nieto-Morales, Peeters, and Lotta Reference Nieto-Morales, Peeters and Lotta2024), it also provides them with the independence needed to resist corrupt practices without fear of retaliation (Dahlstrom, Lapuente, and Teorell Reference Dahlstrom, Lapuente and Teorell2012; Meyer-Sahling, Mikkelsen, and Schuster Reference Meyer-Sahling, Sass Mikkelsen and Schuster2018). Empirical evidence suggests that, more often than not, corruption is driven by heightened political control rather than by bureaucratic autonomy.

The protection of the national interest follows a similar logic. Formally, elected officials have the legitimate power to design and implement policies on behalf of the population. Still, they may use this legitimacy to benefit cronies rather than the general public (Haber Reference Haber2013). Worse, they may aggrandize their own powers, therefore weakening the democratic system and produce democratic backsliding (Bermeo Reference Bermeo2016).

In one of his seminal articles, O’Donnell (Reference O’Donnell1994) argued that one of the perils of the third wave of democratization was that some of the newly minted regimes stagnated at the “delegative democracy” stage. This meant that elected officials had substantive powers to govern without any accountability throughout their terms. In such cases, it is hard to claim that electoral mandates are sufficient to assure that public officials are governing based on the national interest when they are not held accountable by checks and balances.

In these cases, public employees may use their discretion (statutory or not) to influence policy outcomes. As proposed by Yesilkagit et al. (Reference Yesilkagit, Bauer, Guy Peters and Pierre2024, 415), they should follow “a conception of the state that is not neutral in the classic liberal sense, but protective of liberal principles of governance.” The challenge here is that one should expect that bureaucrats would hold values more closely associated with the national interest (in this case, the liberal democracy) than those held by politicians – an assumption often dismissed by, among others, Public Choice scholars (Tullock Reference Tullock and Rowley2005; Tullock, Seldon, and Brady Reference Tullock, Seldon and Brady2002). We, thus, echo Yesilkagit et al.’s (Reference Yesilkagit, Bauer, Guy Peters and Pierre2024) proposition that we must strengthen what they call state guardianship that function as additional checks and balances against illiberal politicians.

2.2 Bureaucratic Politics

In considering the autonomy of the civil service, we should remember that civil servants, and government organizations, are not necessarily passive actors waiting for the rules to be defined by others (Dahlström and Lapuente Reference Dahlström and Lapuente2022; Peters Reference Peters2018). Rather, they may be interested in carving out some greater scope for independent action (Downs Reference Downs1965; Gailmard and Patty Reference Gailmard and Patty2007; Guedes-Neto Reference Guedes-Neto2023). Whether done overtly or by stealth, there is a continuing political struggle within government, a form of bureaucratic politics, that attempts to clarify the rules of the game for accountability, autonomy, and policymaking (see Bach and Wegrich Reference Bach, Wegrich, Andeweg, Elgie, Helms, Kaarbo and Müller-Rommel2020; Peters Reference Peters and Lane1987).

The autonomy of bureaucracy, and of individual public servants, is one of the classic questions in the comparative study of public administration. This is generally not phrased in terms of autonomy per se but more often in terms of the extent to which political leaders are interested in, and capable of, controlling the public service. The other way in which this question can be seen is in the extent to which public servants attempt to pursue a policy agenda of their own, as opposed to being willing to accept the directions of their political leaders (Aberbach, Putnam, and Rockman Reference Aberbach, Putnam and Rockman1981).

The political conflict between politicians and civil servants can be seen through a number of lenses. One has been the traditional lens provided by Max Weber (Reference Weber, Roth and Wittich1978) and Woodrow Wilson (Wilson Reference Wilson1887). Others have sought to develop a taxonomy of the types of interactions (Peters Reference Peters and Lane1987). Others have conceptualized the relationship between these actors as a bargain, with public servants ceding direct control over policy in exchange for less overt influence over policy, as well as permanent appointments (Hood and Lodge Reference Hood and Lodge2006). Still, others have examined the capacity of public servants to act as policy entrepreneurs and to be a more active participant in the policy process (Coyne and Hall Reference Coyne and Hall2018; Silveira, Cohen, and Lotta Reference Silveira, Cohen and Lotta2024; Zhang, Zhao, and Dong Reference Zhang, Zhao and Dong2021).

While this literature on politicians and bureaucrats began with the ideas of Weber and Wilson, it has evolved to accept that public servants have, and indeed should have, a good deal more autonomy from the control of politicians. Some of this shift has reflected the ideas of “empowering” public servants to make more of their own decisions (Peters and Pierre Reference Peters and Pierre2000) as a means of improving the morale and performance of these public employees. In addition, there has been a greater understanding of the importance of allowing civil servants to make more decisions simply because they have more information, especially about applying law to individual cases at the street level.Footnote ²

The study of politicization of the bureaucracy has been the logical extension of the study of relationships between politicians and their public servants. Political leaders rather naturally want their civil servants not to be autonomous, but rather to be committed to their same goals as the government of the day. Therefore, whether by attempting to control who becomes or remains a public servant, or by exerting control over their behaviors while in office, there have been attempts to control bureaucracy (Neuhold Reference Neuhold2014; Peters and Pierre Reference Peters and Pierre2004).

2.3 The Illiberal Era and Bureaucratic Autonomy

This debate about the autonomy of public servants has become more significant during a period of democratic backsliding and increasing politicization of the bureaucracy. As illiberal political leaders assume office, questions of the autonomy of bureaucrats to make their own decisions are crucial for the quality of governance and democracy (Bauer et al. Reference Bauer, Guy Peters, Pierre, Yesilkagit and Becker2021; Lotta, Tavares, and Story Reference Lotta, Tavares and Story2024; Moynihan Reference Moynihan2022; Moynihan and Roberts Reference Moynihan and Roberts2021; Peters and Pierre Reference Peters and Pierre2019; Story, Lotta, and Tavares Reference Story, Lotta and Tavares2023). Given that the policies favored by these illiberal leaders have frequently been questionable legally, and may violate fundamental civil liberties, individual civil servants may have to make their own decisions about whether to exercise some autonomy or to accept directions from the new political leaders.

The attempts to politicize the bureaucracy have become extremely intense and, in some cases, successful, in the illiberal era. For example, Victor Orbán in Hungary has gained a very high level of control over the bureaucracy there (Hajnal and Boda Reference Hajnal, Boda, Bauer, Peters, Pierre, Yesilkagit and Becker2021), and the Law and Justice Party had in Poland until defeated in the election of 2023. During his first term in power, Donald Trump attempted to politicize the American bureaucracy by dismissing all members who did not agree with his policies, and has promised to extend his control if he is reelected (Moynihan Reference Moynihan2022; Moynihan and Roberts Reference Moynihan and Roberts2021). In fact, when Trump returned to the presidency in 2025, many of his first executive orders demonstrated his willingness to fulfill his promise.

An autonomous and capable bureaucracy appears to threaten the agendas of illiberal leaders – and those leaders seem to be recognizing that. On February 7th, 2025, The Washington Post detailed how federal bureaucrats were resisting Trump’s orders (Davies Reference Davies2025). In the article, Jocelyn Samuels, a former commissioner at the Equal Employment Opportunity Commission (EEOC), stated, “If I wasn’t going to use my position to speak out for what I think is right as a matter of law and policy, I might as well not occupy the position at all.” On January 28th, Government Executive reported that she had been fired (Newhouse Reference Newhouse2025).

The question of the autonomy of civil servants to make decisions under illiberal regimes leads to two familiar trichotomies. The first is Albert Hirschman’s (Reference Hirschman1970) idea of exit, voice, and loyalty. Does the civil servant who does not agree with current policies leave the government, does he or she complain about the policies, or do they merely remain and do their job? There is some evidence of civil servants leaving the government under leaders such as Donald Trump (Guedes-Neto Reference Guedes-Neto2022; Moynihan and Roberts Reference Moynihan and Roberts2021), Jair Bolsonaro (Sá e Silva Reference Sá e Silva2020), and Viktor Orbán (Hajnal and Boda Reference Hajnal, Boda, Bauer, Peters, Pierre, Yesilkagit and Becker2021), but most have chosen to remain.

For those who choose to remain in the civil service during an illiberal regime, there is yet another trichotomy of choices about how they will perform their duties: working, shirking, and sabotaging (Brehm and Gates Reference Brehm and Gates1999). Civil servants can choose to work as they have throughout their careers (faithfully following directives from their superiors), they can choose to shirk their duties and do as little as possible for programs with which they disagree, or they can attempt to sabotage the regime’s programs (Schuster et al. Reference Schuster, Sass Mikkelsen, Correa and Meyer-Sahling2021). The latter two choices involve the civil servant trying to evade accountability to elected officials, and pursuing their own conceptions of what constitutes appropriate government action.

Both shirking and sabotage can be seen as a form of voice in Hirschman’s trichotomy. One involves the civil servant displaying discontent with the regime by minimizing his or her contribution to the mission of the agency.Footnote ³ Sabotage is a more extreme behavior, involving undermining in some way the performance of the organization and the delivery of a program in which the civil servant does not believe. That lack of support may come from a policy disagreement, or in the illiberal era it may arise because the civil servant believes the policy undermine basic constitutional principles (Yesilkagit et al. Reference Yesilkagit, Bauer, Guy Peters and Pierre2024).

2.4 Bureaucratic Resistance

The willingness of public servants to resist commands from their political superiors and to pursue goals that they deem more appropriate has been discussed several ways in the public administration literature Rosemary O’Leary (Reference O’Leary2019) has discussed this behavior as a “guerilla government” and Jan Olsson (Reference Olsson2016) has described it as “institutional subversion.” At an organizational level, this resistance to controls can be considered as a search for organizational autonomy and the pursuit of organizational goals (perhaps mandated by law or democratic norms) rather than accepting instructions from politicians. Further, career civil servants may believe that they better represent those organizational goals, and perhaps even the goals of government as a whole, than do the short-term inhabitants of ministerial positions.

Resistance may occur for different reasons and through different channels. Hollibaugh Jr., Miles, and Newswander (Reference Hollibaugh, Gary, Matthew and Newswander2020) fielded a conjoint experiment with American bureaucrats to demonstrate that policy disagreements, ethical concerns, and the policy’s negative impact influence their willingness to disobey their principal or engage in confrontation, delay compliance, or whistle-blow.

In addition to this, the literature shows that some bureaucrats will actively mobilize support from external actors (Milhorance Reference Milhorance2022; Milhorance et al. Reference Milhorance, de Andrade, le Coq, Sabourin, Morais de Sá e Silva and Gomide2024; Waisbich Reference Waisbich2024), such as nonprofits and international organizations, to safeguard their policy priorities or, at least, to avoid policy dismantling. While these studies focused on bureaucratic resistance during the illiberal administration of Jair Bolsonaro in Brazil, there is also evidence that public employees will mobilize external actors to advance their policy preferences in normal times, even when these may not be aligned with the priorities of elected officials (Abers Reference Abers2019; Abers and Tatagiba Reference Abers, Tatagiba, Rossi and von Bülow2015; Rich Reference Rich2013).

As an example, consider Donald Trump’s executive order “Restoring Accountability to Policy-Influencing Positions within the Federal Workforce” (January 20th, 2025). On his first day in office for the second term, the president explicitly wrote that “In recent years, however, there have been numerous and well-documented cases of career Federal employees resisting and undermining the policies and directives of their executive leadership.” The executive order, then, acknowledged the existence of bureaucratic resistance and aimed at recreating the Schedule F career (see Moynihan Reference Moynihan2022), which grants the presidency increased powers to hire and fire at will. Other presidential acts followed the same trend. That is, for Trump, the issue was relevant enough to merit such responses.

We build our analysis of the possible types of resistance of bureaucrats on the study of Brehm and Gates (Reference Brehm and Gates1999) and their trichotomy of “working, shirking, and sabotage.” The logic of this analytic scheme is that most civil servants will perform their assigned tasks most of the time (Pierre and Peters Reference Pierre and Guy Peters2017). However, when confronted with negative working situations they can either shirk, or do as little as possible, or actually sabotage the work of the organization.Footnote ⁴ These latter two options, and especially sabotage, involve their violating the usual terms of their employment, and therefore represent significant career choices for them.

We agree with Perry and Wise (Reference Perry and Recascino Wise1990) that, in general, public employees will hold a “public service motivation.” This means that they opted-in for the public administration due to their willingness to serve the population. We assume that, at least in democratic contexts, this means that civil servants will attempt to preserve their public organizations as much as possible based on the perception that they provide a desirable service to the population. In this sense, new leaders attempting to impose an illiberal policy would be going against their interests and the interests of the population they serve. Therefore,

The willingness of civil servants to resist commands from their political superiors may be a function of several personal and organizational factors – including, but not limited to, their democratic values and policy preferences. We, however, opt for discussing a less obvious variable. Consider perceived job status. Resistance comes at a cost. We have previously mentioned the case of Jocelyn Samuels, the EEOC commissioner fired by Trump. In some cases, illiberal leaders will remove their opponents from the public administration. In other cases, when job stability is more protected, illiberal politicians and their allies will engage on abusive supervision (Story, Lotta, and Tavares Reference Story, Lotta and Tavares2023), sidelining (Muno and Briceño Reference Muno, Briceño, Bauer, Peters, Pierre, Yesilkagit and Becker2021), or even political persecution (Hajnal and Boda Reference Hajnal, Boda, Bauer, Peters, Pierre, Yesilkagit and Becker2021). It is natural that before resisting, public employees will calculate the costs of their actions.

When civil servants believe that they could easily find a well-paying job in the private sector, the fear of being fired does not represent a high cost – at least not as high if compared to those who believe they could not find a good job in the private sector. To corroborate, Selden and Moynihan (Reference Selden and Moynihan2000) find that bureaucrats are less likely to quit the higher their salary is. In such cases, the risk of becoming unemployed would likely demotivate some civil servants to defy their bosses’ preferences.

Llorens and Stazyk (Reference Llorens and Stazyk2011) argue that public–private salary equity is not a strong predictor of turnover. Even if that is the case, the perception of better opportunities should still matter for decisions to shirk or sabotage. Those who do not perceive the difference between public and private salaries as important may have opted to work for the government in the first instance not due to income but because of their public sector motivation (Perry Reference Perry2000; Perry and Wise Reference Perry and Recascino Wise1990) and the ability to help society. To a certain extent, this resembles what de Graaf (Reference de Graaf2011) calls the “by-the-book professionals.” The label regards those public employees who are not loyal to the stakeholders of their organization, but to society and its citizens. In his own words, “blindly following the minister is not something a professional [of this kind] should do” (de Graaf Reference de Graaf2011, 295).

Following these two rationales, we hypothesize that:

Our next proposition follows from the previous. Fear of punishment for shirking or sabotage should be the starkest among those who do not enjoy job stability (Tullock Reference Tullock and Rowley2005). While the rational choice approach may justify this proposition based on the employees’ self-interest of preserving their jobs, there is also the issue of political control. Cornell and Lapuente (Reference Cornell and Lapuente2014) propose that in meritocratic administrations, where most civil servants enjoy job stability, democracy tends to be more stable given the constraints imposed on political action (see also Cornell and Grimes Reference Cornell and Grimes2015) taken by superiors to punish subordinates.

This argument is at the root of Wilson’s proposition of separating politics and administration. In fact, he concludes his seminal work with a series of questions, one of them being:

This summarizes his proposition that a merit-based administration, insulated from politics, should not serve solely at the will of a political superior, but also in response to the public interest and its own conscience. If, however, the spoils system prevails, public officers must obey their superior’s orders alone to preserve their employment.

To be sure, the merit system does not eliminate the politics from the bureaucracy. Civil servants have their own preferences, motivations, and agendas. However, by assuring meritocratic recruitment and retention, bureaucrats need not to align their politics with the politics of their superiors. This allows tenured bureaucrats to disobey elected officials when faced with illiberal expectations. Of course, this also comes with the cost that resistance may occur even when elected officials attempt to implement policies that are not illiberal.

Even though the merit system has become the norm in most democratic countries, not all public employees are tenured. Some enter the government through political appointments, others are hired through temporary contracts, and so on. Nontenured employees will surely face higher costs of resistance than tenured bureaucrats. This should influence their decision to resist illiberal orders. Thus, we expect that tenure operates as a constraint against political control over the bureaucracy since it grants career bureaucrats more autonomy to resist illiberal policies with fewer risks of facing punishment than untenured public employees would.Footnote ⁵

We hypothesize that:

Finally, we hypothesize that one factor affecting resistance by bureaucrats is whether the individuals themselves are supervisors. The closer a public employee is to the principal (in this case the political leader), the harder it is for him or her to resist. Tullock (Reference Tullock and Rowley2005) argues that bureaucrats interested in career advancement will opt for following the orders of their superiors. Aberbach and Rockman (Reference Aberbach and Rockman1995) find support for this proposition at least among public employees who become loyalists to new administrations regardless of their individual preferences.

While supervisors may enjoy more formal autonomy in directing their teams, their positions are often contingent upon maintaining alignment with political leaders. If they resist political directives or fail to adequately implement political priorities, they face removal from leadership roles. This dynamic creates a powerful incentive for compliance, despite their nominal autonomy. The need to secure their position within the bureaucracy often outweighs their ability to act independently. As a result, supervisors may be less willing to resist undemocratic policies compared to lower-level bureaucrats, who are further removed from direct political oversight. This reinforces the argument that proximity to political actors can constrain resistance, even among those with greater managerial discretion.

This leads to our last hypothesis:

This proposition is also close to another ideal type of top manager proposed by de Graaf (Reference de Graaf2011): the “society’s neutral servants.” They are the “most loyal to their ministers and have the fewest conflicts with them” (de Graaf Reference de Graaf2011, 296). These managers do so not necessarily act in this manner because of possible personal gains, but because they believe they are serving society when following orders. Not all supervisors will indeed be like that, as pointed out earlier (see de Graaf Reference de Graaf2011). However, we propose that the higher someone is in the hierarchy, the closer they will be to political actors and more willing to take direction from those politicians, whether for the sake of greater benefits or their willingness to implement the orders of their principals.

We should remember, however, that those in the upper echelons of the bureaucracy will be more experienced and more knowledgeable about policy, and therefore may be more likely to have ideas of their own. The reactions of highly placed bureaucrats therefore may be substantive disagreements with their ministers, or with political advisors. The job of the civil servant, in addition to following the directions of their political superiors, is to provide “frank and fearless” advice – something they cannot do if they only tell their superiors what they want to hear.

2.5 Summary

This section has discussed two trichotomies that can describe the behaviors of civil servants when confronted with policy decisions with which they do not agree taken by their political leaders. While the majority of civil servants will be loyal to their organizations and do their work as expected, there are situations in which some will seek to leave the organization or will not work as diligently as they might normally. In the extreme, there are times in which some civil servants may decide to attempt to undermine the work of the organization, especially when they receive orders that they consider illegal or immoral. The likelihood that civil servants will engage in shirking is, we argue, increased by illiberal leaders who seek to undermine liberal democratic values.

This section has developed a series of hypotheses about the behavior of civil servants when faced with leadership within their organizations that may be advancing illiberal proposals. Although much of the bureaucratic autonomy literature emphasizes the organizational basis of autonomy, we are focusing our research on the individual level, and the hypotheses assess the extent to which the characteristics of individual civil servants may affect their behavior vis-à-vis illiberal political leadership. This is an enduring question in public bureaucracies, but has become especially relevant with the growth of populist, illiberal political control of governments in many area of the world. Section 3 will test the hypotheses we have developed and provide insights into the ways in which civil servants cope with changes in their working environments.

4.1 List Experiments

Shirking and sabotage by civil servants are often deemed undesirable behaviors in normative terms (Pierre and Peters Reference Pierre and Guy Peters2017). Public workers generally have been observed preferring to do at least some work rather than none. This is especially true if we consider the findings of research on public service motivation (see Perry Reference Perry1997) that indicates that in most settings civil servants are committed to their jobs Furthermore, in many cases, not working may go against the bureaucrat’s professional code of conduct, therefore risking different kinds of punishment: a lower salary, in agencies that adopt a pay-for-performance system, fewer chances for promotions, or even the employee’s dismissal. These multiple possibilities will likely produce a social desirability bias in any survey responses related to working and sabotage.

Given the potential lack of sincerity in respondents’ self-reporting regarding the consideration of adopting such strategies, direct questioning is deemed unlikely to produce reliable results. To mitigate social desirability bias, we employ a list experiment, as recommended by Peters and Guedes-Neto (Reference Peters, Guedes-Neto, Vigoda-Gadot and Vashdi2020), which allows subjects to avoid directly admitting to holding controversial positions or behaviors while still enabling researchers to draw inferences about such behaviors and attitudes. Several authors validated the benefits of list experiments in contexts of high social desirability bias (Gonzales-Ocantos et al. Reference Gonzales-Ocantos, Jonge, Meléndez, Osorio and Nickerson2012; Oliveros Reference Oliveros2016). Gonzalez-Ocantos et al. (Reference Gonzales-Ocantos, Jonge, Meléndez, Osorio and Nickerson2012), for example, demonstrate that voters will lie if asked directly about vote-buying but will be sincere if the question is part of a list experiment.

The list experiment, also known as the item count technique, was originally introduced by Miller (Reference Miller1984) in her doctoral dissertation as a method to elicit truthful responses to sensitive survey questions while minimizing social desirability bias. Since then, it has become a widely used and well-validated method in political science and survey research. A meta-analysis by Blair, Coppock, and Moor (Reference Blair, Coppock and Moor2020) reviewed nearly 300 studies spanning 30 years and confirmed that list experiments provide a robust and reliable means of estimating the prevalence of sensitive behaviors or attitudes, particularly when the underlying assumptions – such as no design effects and truthful reporting – hold.

The logic of the list experiment is straightforward. Respondents are randomly assigned to either a control or treatment condition. Those in the control group receive a list of four nonsensitive items and are asked to indicate how many of those items apply to them, without specifying which ones. The treatment group, in contrast, receives the same four items with the addition of a fifth, potentially sensitive item. Respondents in both groups report only the total number of items they endorse, preserving anonymity regarding their response to the sensitive item. Because random assignment ensures that any preexisting differences between the groups are distributed equally, the difference in the mean number of endorsed items between the treatment and control groups provides an unbiased estimate of the proportion of respondents selecting the sensitive item (Sniderman et al. Reference Sniderman, Druckman, Green, Kuklinski, Lupia, Druckman, Greene and Kuklinski2011).

This methodology is particularly useful in contexts where respondents may be reluctant to disclose sensitive views or behaviors due to fear of judgment, political pressure, or legal concerns. Unlike direct questioning, which can introduce sensitivity bias, the list experiment mitigates this risk by making it impossible to determine an individual respondent’s answer to the sensitive question, thus encouraging more honest responses. Furthermore, the difference-in-means estimator used in list experiments has been extensively validated in the literature, demonstrating its robustness even when response ranges differ between conditions.

In our initial list experiment, we regard shirking as a proxy for resistance. Subjects allocated to the control condition encounter the following vignette:

In the treatment condition, participants received the same text as the control group, supplemented with a fifth item: “it reduces citizens’ political rights, such as the freedom of expression or press.” Subsequently, all participants were prompted with the question: “In your opinion, how many of the previous scenarios would lead a standard civil servant to dedicate fewer efforts to the project in comparison to other activities? Please, answer only with the number of scenarios.” Response options ranged from 0 to 4 in the control condition and 0 to 5 in the treatment condition.

Three theoretical considerations guided our experimental design. Firstly, to further alleviate social desirability bias associated with shirking, respondents were encouraged to consider a “standard civil servant.” This approach, as suggested by Mutz (Reference Mutz2011), allows subjects to express their perceptions without directly admitting to potentially politically incorrect behaviors.

Secondly, to mitigate measurement error arising from extreme responses (0 or 4), we carefully selected control items with negative correlations, ensuring that respondents would be less likely to endorse all or none of the listed options. For instance, it is improbable for someone selecting “is very similar to every other project that she/he has always worked at” to also choose “is entirely new to her/him, requiring training and additional efforts.” This structuring reduces the likelihood of flooring and ceiling effects, which can bias estimates by artificially constraining variation in responses (Glynn Reference Glynn2013).

By implementing this strategy effectively, we anticipated an average response in the control condition of approximately 2, meaning that respondents would endorse some – but not all – items, thereby improving the precision of our estimate of the sensitive item’s prevalence. A well-balanced distribution of responses ensures that the difference-in-means estimator remains unbiased, as extreme response clustering (all 0s or all 4s) could otherwise distort the estimation of treatment effects.

Thirdly, to further prevent flooring effects, we included at least one high-likelihood item, namely “creates a political advantage to groups that she/he is against.” Given the nature of political dynamics in bureaucratic environments, we anticipated that this item would receive broad agreement among respondents, thereby reducing the likelihood of zero endorsements across the entire control group. This design choice aligns with Kuklinski et al. (Reference Kuklinski, Cobb and Gilens1997), who demonstrated that including at least one widely endorsed item stabilizes variance and improves estimator reliability. Together, these methodological choices ensure that our list experiment adheres to best practices, maximizing the accuracy of our treatment effect estimates while minimizing distortions caused by extreme response patterns.

Our subsequent experiment, focusing on the possibility of sabotage, mirrored the previous one. The options and survey questions remained unchanged, while the introductory text read as follows:

To summarize the key differences, while both experiments investigate bureaucratic resistance, they capture distinct forms of noncompliance. The first experiment, centered on shirking, examines bureaucrats who choose to withdraw effort or passively disengage from implementing a project assigned to them. This is reflected in the introductory text, which emphasizes civil servants who neglect their responsibilities rather than actively obstructing a policy. In contrast, the second experiment focuses on sabotage, a more deliberate and proactive form of resistance, where bureaucrats intentionally undermine a project to ensure its failure. The introductory text in this case explicitly frames the scenario as one in which public employees work against a project rather than merely failing to contribute to its execution. By maintaining consistent response options while altering the framing of resistance, the design allows us to compare bureaucratic behaviors along a continuum from passive to active resistance.

Following the data collection phase, we conducted data analysis using two-way t-tests to compare differences in means. Our underlying assumption, once again, is that the randomization process effectively neutralized any effects that could be attributed to other covariates (Peters and Guedes-Neto Reference Peters, Guedes-Neto, Vigoda-Gadot and Vashdi2020). Additionally, we performed post-hoc power analyses to evaluate the statistical reliability of our findings, ensuring that only results with a minimum power of 0.80 were considered robust enough for interpretation (Bausell and Li Reference Bausell and Yu-Fang2006). While a priori power analysis is crucial in experimental designs to determine the required sample size beforehand, post-hoc power analysis serves an equally important role in studies with fixed or convenience samples, where researchers must assess whether observed effects are adequately powered (Perugini, Gallucci, and Costantini Reference Perugini, Gallucci and Costantini2018). This approach allows us to filter out potentially underpowered results, reducing the risk of Type II errors and ensuring that our conclusions are drawn from statistically meaningful effects.

The balance table is presented in Table 2. Notably, the only variable that is not balanced across conditions is discretion. To address this and account for other potential omitted variables, we conducted a series of robustness checks as we further explain in the following sections.

4.2 Heterogeneous Treatment Effects

Our theoretical framework requires the assessment of heterogeneous treatment effects, indicating experimental outcomes that vary based on respondent characteristics. Initially, we explore variations in treatment effects by country, dividing the sample into three groups: the United States, United Kingdom, and Brazil, and conducting individual two-way t-tests for each case. Again, we are cautious with cross-country comparisons because our samples are not nationally representative of each country’s bureaucracy.

Subsequently, we investigate treatment effects based on respondents’ perceptions of opportunities in the private sector. Our H₂ posited that those civil servants who believe that they would be (financially) better off in the private sector will be more likely to resist undemocratic policies. Respondents were prompted to envision a scenario where a civil servant, possessing equivalent experience and contacts to their own, opts to seek employment in the private sector. Among the 942 respondents, 66.56% anticipated a higher salary in the private sector, 12.10% expected the same salary, and the remaining 21.34% anticipated a lower salary. To test H₂, we dichotomize responses into two categories: (1) anticipation of a higher salary in the private sector and (2) expectation of the same or lower salary.

For testing H₃ (the effect of tenure) and H₄ (the effects of supervisory responsibilities), our categorization approach is more straightforward. We differentiate between respondents who are tenured and those who are not for H₃, and for H₄, we distinguish between supervisors and subordinates. Given our lack of hypothesized country differences in heterogeneous treatment effects for these individual-level categorizations, we solely test them based on the entire sample. This methodological decision is also influenced by our sample size, which would be substantially reduced if we attempted to select, for instance, only tenured individuals from specific countries.

5.1 H₁: Are Bureaucrats Willing to Shirk and Sabotage?

We present our baseline results in Tables 3 and 4. Before interpreting the treatment effects, we emphasize three key points warranting special attention. Firstly, the average response in the control condition closely approximates two, validating our empirical design. Secondly, across all cases, the p-value is below 0.05 and the statistical power exceeds 0.80. As per professional standards, these thresholds signify statistical significance in the difference of means. Thirdly, as previously explained, our focus lies on the disparity of means between the control and treatment conditions, indicating respondents who selected the fifth statement unique to the treatment condition (i.e., the undemocratic scenario).

While we anticipate resistance among bureaucrats regardless of country (H₁), we expect to observe heightened intent to resist in Brazil, followed by the United Kingdom and the United States (in this sequence). Our findings corroborate this hypothesis, supporting H₁ and aligning with our comparative expectations. American civil servants exhibit the lowest inclination towards shirking or sabotage, yet they still demonstrate a notable willingness to resist. Regarding shirking (Table 3), 58.8% of US respondents assert that a standard civil servant would consider shirking if tasked with implementing a policy contrary to basic democratic rights. This sentiment is even more pronounced among British (65.2%) and Brazilian (84.4%) bureaucrats. Comparable results are observed for sabotage: 59.9% among American respondents compared to 69.4% among those in the UK sample and the whole treated sample for Brazil.

The difference between the treatment effects of the US and UK samples is less than 0.10 in both instances. Furthermore, they never surpass the 70% mark. These figures noticeably contrast with the outcomes observed in the experiments conducted in Brazil. In this developing country, 84.4% of respondents believe their peers would be inclined to shirk if tasked with implementing undemocratic policies, and even more strikingly, 104.0% indicate a potential willingness to sabotage the government. The latter result surpasses the 1.00 mark, suggesting a triggering effect that prompts respondents to demonstrate a readiness to engage in sabotage in other contexts as well when exposed to the undemocratic frame.

5.2 H₂: Does the Perception of Better Opportunities in the Private Sector Affect Resistance?

Next, we examine the influence of perceived opportunities in the private sector on the willingness to resist undemocratic policies. We hypothesize that individuals who perceive better opportunities in the private sector will demonstrate a greater propensity to shirk and sabotage compared to others. Our data substantiate this hypothesis. As depicted in Table 5, 69.6% of respondents who anticipate better salaries in the private sector also believe their peers would shirk if tasked with implementing an undemocratic policy. In contrast, this figure stands at 55.1% among those who expect similar or inferior earnings in the private sector.

Similarly, the list experiment investigating sabotage as a form of resistance yields analogous results. Among those perceiving superior opportunities in the private sector, 71.4% believe that standard civil servants would engage in sabotage when confronted with policies undermining freedom of expression and the press. Conversely, only 64.5% of those not anticipating higher salaries in the private sector hold the same expectation.

To reaffirm, all treatment effects exhibit considerable magnitude. In each scenario, over half of the treated individuals expressed their intent to shirk or sabotage if tasked with implementing an undemocratic policy. This holds true regardless of whether individuals perceive themselves as underpaid or overpaid. However, there is a notable escalation in treatment effects among those anticipating higher earnings in the private sector.

5.3 H₃: Do Tenured Bureaucrats Differ from Untenured Public Employees?

Our subsequent hypothesis posits that tenured public employees are more inclined to shirk and sabotage compared to noncareer bureaucrats. However, our findings fail to support this assertion. Table 6 illustrates the results for shirking, revealing virtually identical differences in means for both tenured (0.649) and untenured civil servants (0.640). Contrary to our hypothesis, there appears to be no significant disparity in propensity to shirk between these two groups.

Additionally, we present the heterogeneous treatment effects of the sabotage experiment within the same table. Surprisingly, the results contradict our hypothesis; 71.9% of the untenured respondents believe their peers would be inclined to sabotage, whereas this sentiment is shared by only 66.0% of the tenured portion of the sample. This unexpected finding suggests that tenure status may not be a decisive factor in predicting willingness to engage in sabotage.

The heightened inclination of untenured bureaucrats to engage in sabotage compared to their tenured counterparts may arise from various factors, including perceived job insecurity, limited exposure to consequences, and a desire to assert dedication to democratic principles. For instance, some untenured public employees might view their tenure in government as transitory, prioritizing the cultivation of a pro-democracy reputation over compromising their democratic values. Additionally, they may be more tolerant of being labeled as saboteurs by colleagues, recognizing their temporary status within the organization as a buffer against the potential repercussions of a tarnished reputation.

Conversely, both tenured and untenured bureaucrats demonstrate similar propensities to shirk, indicating that the influences on shirking behavior may diverge from those affecting sabotage tendencies. These findings underscore the intricate interplay of job security, experience with consequences, and career aspirations in shaping employee conduct within bureaucratic settings, with tenure status exerting distinct effects on the inclination to engage in acts of resistance versus avoidance of work-related duties.

5.4 H₄: Do Supervisors Differ from Subordinates?

Our final hypothesis posits that subordinates will demonstrate a greater willingness to resist than their supervisors. The findings from the combined datasets substantiate our hypothesis at standard levels of significance and power, as shown in Table 7. Specifically, our analysis reveals that 60.9% of supervisors believe that a standard civil servant would shirk in the face of an undemocratic policy. This treatment effect escalates to 67.9% among subordinates alone.

The disparity between groups is even more pronounced for sabotage. While 61.6% of supervisors are keen to sabotage, a notable 74.8% of respondents without supervisory status expect this form of democratic resistance. Thus, our results consistently support the hypothesis that subordinates exhibit a greater propensity to resist compared to their superiors in both scenarios.

It’s noteworthy that while the disparity in treatment effects is 7 percentage points in the shirking experiment, it widens to 13.2 percentage points when respondents are queried about sabotage. This indicates that closeness to the agency’s leadership acts as a more substantial deterrent to sabotage than to shirking. Subordinates demonstrate a greater propensity to opt for the most radical alternative – sabotage – compared to shirking, such as deliberately missing deadlines. In essence, bureaucrats who maintain a distance from political leaders exhibit a clear willingness to confront undemocratic officials.

5.5 Does Peer Pressure Affect Bureaucratic Response?

We crafted the list experiment to minimize the influence of social desirability bias, ensuring respondents could express their willingness to resist undemocratic policies freely. Nevertheless, real-life settings entail inherent costs. For instance, Pierre and Peters (Reference Pierre and Guy Peters2017) assert that peer pressure often compels bureaucrats to prioritize work over shirking. To address this, we posed a follow-up question to subjects from the United States and the United Kingdom: “Would your answer have been the same if you knew that most work colleagues would disagree with what you answered?” Responses ranged from “It would definitely change” (0) to “It would definitely remain the same” (10). Our results reveal that the majority of public employees perceive minimal influence from peer pressure, as depicted in the histogram plotted in Figure 2.

Figure 2 Histogram: Would the respondent’s answer remain the same if colleagues disagree?

Figure 2Long description

The x-axis represents a scale from 0 (indicating the respondent would change their answer) to 10 (indicating they would maintain their answer). The y-axis represents the percentage of respondents selecting each option, from 0 to 50. The distribution is skewed toward higher values, with increasing frequency as the scale progresses. There are 10 data points, and moving from 0 (will change answer) to 10 (will maintain answer), the values are: 1 percent, 0.5 percent, 1 percent, 4 percent, 4.5 percent, 8 percent, 5 percent, 9 percent, 17 percent, and 50 percent. This suggests hta most people would not be willing to change their response.

Analyzing our sample, we observed that 38.21% of respondents chose 10, and 68.31% selected at least 8 (i.e., from 10 to 8), indicating a prevailing perception of low costs associated with peer pressure. None of the other categories (7 and below) was chosen by more than 9% of the respondents. Since there seems to be a significant difference between the subsamples choosing 8 or more on one side, and 7 or less on the other, we created a dummy variable categorizing respondents into these two groups. We refer to the first group (8–10) as those who perceive low costs of peer pressure, while the second (0–7) perceives high costs. We then conducted two-way t-tests to assess the treatment effects among each of these groups.

The treatment effects of the shirking experiment are presented in Table 8. We find that even among those who would not be willing to change their response (low-cost perception), there is still a significant number of bureaucrats who would be willing to shirk. The treatment effects in the low perception category are 54% in the United States and 67% in the United Kingdom. Based on these findings, we propose a conservative estimation of the share of bureaucrats that would be willing to shirk regardless of peer pressure (at least in our convenience sample). In the United States, these are the 54% identified in the list experiment among the 74.76% of the sample, which is in the low perception category, yielding 40.37% $[0.54\times 0.7476=0.4037]$ of the American respondents. In the United Kingdom, this is 67% of 61.42% of the sample, that is, 41.15% $[0.67\times 0.6142=0.4115]$ .

To recap, our most flexible estimation, which considers only treatment effects among those with a low-cost perception of peer pressure, supports our cross-country expectation that British bureaucrats (67%) are more willing to resist than those from the United States (54%). Our conservative estimation, which accounts for the share of respondents who see low costs in peer pressure, produces only a very small difference across countries: 41.2% (United Kingdom) and 40.4% (United States).

We present the results for sabotage in Table 9. We find that among those who are not responsive to peer pressure, this starker type of resistance is even more likely to exist both in the United States (55%) and the United Kingdom (82%). Following the same strategy proposed in the previous paragraph, we estimate that 41.12% $[0.55\times 0.7476=0.4112]$ of the American public employees who participated in our experiment believe that sabotage would take place in case of implementing an undemocratic policy. This compares to 50.36% $[0.82\times 0.6142=0.5036]$ of the British sample. These estimations support our cross-national that it is more likely to identify resistance in the United Kingdom than in the United States regardless of the measurement strategy.

In addition to the cross-national comparisons, even our most conservative estimations yield a compelling result: a substantive share of participants demonstrates a readiness to resist undemocratic policies. This finding not only reinforces H₁ but also underscores the pivotal role of bureaucrats as core veto actors within democratic backsliding processes. It emphasizes the potential influence wielded by civil servants in safeguarding democratic principles and resisting illiberal leaders, thereby highlighting the critical importance of understanding and engaging with bureaucratic perspectives in efforts to uphold and fortify democratic institutions.

5.6 Vignette Experiment

In addition to our list experiment, we employed a second strategy to address any potentially unrealistic perceptions of decision costs among respondents. Following their response to a series of initial questions, participants were randomly assigned to one of two groups and exposed to a vignette experiment. To ensure unbiased results, we implemented two key measures. Firstly, we exposed participants to a series of survey questions between the list experiment and the vignette experiment. These functioned as confounders, thus minimizing any potential bias. Secondly, we re-randomized all subjects upon entry into the vignette experiment, ensuring an equal chance of assignment to either the control or treatment conditions. The vignette presented to participants was structured as follows:

While a vignette experiment inherently carries a higher risk of inducing social desirability bias compared to a list experiment (Peters and Guedes-Neto Reference Peters, Guedes-Neto, Vigoda-Gadot and Vashdi2020), we incorporated additional measures to introduce a nuanced understanding of decision costs. Firstly, we modified the wording to heighten the perceived stakes: rather than referencing a generic “standard civil servant,” respondents were prompted to consider their own behavior. More importantly, we crafted a vignette that could effectively differentiate between the effects of an undemocratic policy and those of a policy perceived as democratic but personally disliked by the respondent. This deliberate design enables us to pinpoint the specific willingness to resist undemocratic leaders, shedding light on the nuanced dynamics of bureaucratic decision-making under varying policy contexts.

Our results are presented in Table 10. In the pooled sample, we find that those respondents who were treated are, on average, more willing to shirk than those in the control condition at a rate of change of 29.35% $\left[\right(4.085-3.158)/3.158]$ . This finding still supports our expectation that bureaucrats are willing to resist undemocratic policies more than other policies that are undesirable yet democratic.

Here, we also find that the treatment effects are slightly higher among American bureaucrats than among those in the United Kingdom, suggesting either that British respondents were more influenced by social desirability bias or that part of the effects captured in the list experiment were not exclusively related to the undemocratic character of the policy. To be more precise, this suggests that resistance in Britain (compared to the United States) is more likely to occur for reasons that go beyond sustaining democracy.

Furthermore, we do not find any statistically significant effect in the Brazilian sample. This may be caused either by the small sample size or a lack of interest in resisting undemocratic policies compared to policies that Brazilian public employees dislike even though they are democratic. The latter proposition appears to be the case, considering that the average of the control group is considerably higher in Brazil than in all other countries – in the control group, the average in the Brazilian sample is 4.98 compared to 2.80 in the United Kingdom and 2.90 in the United States. That is, even when the policy is not undemocratic, our subjects appear to be more willing to shirk than in the United Kingdom and the United States.

5.7 Omitted Variable Bias and Balance across Countries

In the balance table (see Table 2), we show that our randomization process was successful at assigning similar samples to the two experimental conditions. This allows us to affirm that our experiment is unlikely to have been influenced by omitted variable bias (Mutz Reference Mutz2011), thus justifying the use of two-way t-tests. However, one variable (discretion) was not balanced across groups. On average, those in the treatment group enjoy slightly more autonomy in their work when compared to participants assigned to the control condition.

To account for that and any other effect that does not derive from the treatment, we ran a multivariate OLS regression considering the outcome as the dependent variable, the treatment as the independent variable, and all the covariates collected in the survey (see Table 11). This strategy also helps to mitigate any differences that are caused by the different recruitment strategies employed in the three countries.

Furthermore, we adopted the inverse-probability weights (IPW) technique (see Table 12), a considerably more conservative estimation approach to calculate the average treatment effects. In short, it relies on weighted probabilities to correct missing data (Cattaneo Reference Cattaneo2010). In all cases, they are virtually the same as those presented in Tables 2 and 3 (pooled sample).

We use Imai and Ratkovic’s (Reference Imai and Ratkovic2014) test of overidentifying conditions to assess whether our IPW approach adequately fixed our estimations. We find support for the hypothesis that the treatment models balanced the covariates (p-value < 0.001). In other words, our results do not seem to derive from a lack of balance across treatment and control conditions or differences in the demographic or professional differences of subjects from the three countries.

We adopted the same IPW technique to compare average treatment effects across countries. That is, since the different recruitment approaches led to some differences in the characteristics of respondents from the United States, the United Kingdom, and Brazil, it is important to make sure that any effects are driven by the treatments and country of origin, and not by other covariates (e.g., tenure, political appointment, etc.). Our IPW models by country, which are presented in Table 13, show that our findings are not driven by differences in the sample. While the magnitude of some effects may change, the cross-country order of magnitude is virtually the same.

6.1 Questions for Further Research

Our research has been comparative, but there is a good deal more research that should be done across time and place. One question that will need to be addressed in further research is the extent to which these findings are a function of experience with governments with illiberal tendencies. While we are arguing that Boris Johnson did not meet the usual criteria associated with a populist leader, he still was more like that type of politician than had been previous British prime ministers (Alexandre-Collier Reference Alexandre-Collier2022; Lacatus and Meibauer Reference Lacatus and Meibauer2022). Therefore, all the public servants in these samples have had some experience with leaders who appear to threaten fundamental liberal values in democracy. Do civil servants working is less obviously illiberal political systems have the same willingness to go against the wishes of an elected government?

Civil servants who have not had that experience may not be as sensitized to the threats, and therefore may be less able to conceive of themselves shirking or sabotaging in their workplace. There is evidence, for example, that prior experiences with governments in which the rule of law has been threatened does influence the behavior of officials (Kyriacou and Trivin Reference Kyriacou and Trivin2024). One next step in this type of research is to conduct similar studies in societies in which populism is not a prominent part of the political environment, and to determine if there are similar results.

In a similar vein, it will be interesting to see if civil servants become less willing to engage in shirking and sabotage after the end of an illiberal regime. There have been at least three instances – Brazil, Poland and the United States – in which an illiberal regime has been replaced by elections. Has the experience of illiberal government with its threats to democratic principles altered the ways in which civil servants think about their jobs in a fundamental manner, or is the expressed willingness to engage in those nonconventional behaviors for civil servants situational and dependent upon the nature of the political regime in place at any one time?

It also appears from anecdotal evidence that the experiences of civil servants in illiberal governments have sensitized civil servants in other systems to the potential dangers of this type of behavior within government. Another potentially fruitful line of research is to investigate the extent to which there has been a general transformation of thinking about their roles among civil servants in democratic political systems. While the cases of illiberal governments have been the clearest, the general tendency to increase the politicization of the civil service may be evoking a reaction among public servants who want to defend their autonomy.

Further studies should also attempt to improve data quality, more specifically, the statistical representation of the populations they are targeting. Although they yield statistical power, our samples in Brazil are considerably small, and the convenience samples of public employees from the United States and the United Kingdom come from a large variety of organizations. Since we fielded our first survey in Brazil (see Guedes-Neto and Peters Reference Guedes-Neto, Guy Peters, Bauer, Pierre, Yesilkagit, Peters and Becker2021), it has been replicated by Lobo (Reference Lobo2022) and Pereira (Reference 62Pereira2023) with large samples of Brazilian federal employees, respectively, public prosecutors at the National Treasury and civil servants at the Securities and Exchange Commission. In both cases, the results were fairly similar to ours. Bersch and Lotta (Reference Bersch and Lotta2023) conducted interviews with federal employees at environmental agencies to find similar evidence of willingness to resist. Future studies should consider other agencies and larger sample sizes to confirm ours and these results, as well as to map cross-agency variation in resistance to illiberal policies.

In addition to the previously stated questions, we do not know a great deal about why these civil servants are willing to express intent to depart from the usual expectations of their positions. We have made some inferences about the factors that influence that expression of commitments, based on their positions within organizations and the permanence of their appointments, but we do not have direct evidence about the justifications for their opinions, and the understandings of the civil service that are reflected in their choices. Any number of factors may be associated with the expressed willingness to deviate from the usual expectations of behavior for members of the civil service.

Furthermore, the evidence used in these studies is all about expressed willingness to engage in certain types of behaviors, rather than about actually engaging in those behaviors. The old adage that “talk is cheap” can easily be applied here. It will take a good deal of courage and commitment for a civil servant to risk his or her career over principles, even ones as important as those that may be threatened by contemporary illiberal governments. When faced with the need to decide whether to carry through with these behaviors, what would they really do?

The evidence that is readily available about the decisions of civil servants is that a significant number have been willing to leave government when faced with political leadership they find unpalatable (Ali Reference Ali2019; Bolton, Figueiredo, and Lewis Reference Bolton, Figueiredo and Lewis2020; Dahlström and Holmgren Reference Dahlström and Holmgren2019; Guedes-Neto Reference Guedes-Neto2022; Schuster et al. Reference Schuster, Sass Mikkelsen, Correa and Meyer-Sahling2021). This “exit” behavior appears most obvious in the United States, but has also occurred in other governments with illiberal leaders (Barkat et al. 2024). Exit may be in some ways an easier option, especially for individuals in the upper levels of government who may be able to find new positions in think tanks or consulting rather easily. Remaining in office and attempting to undermine a government that has legitimacy from an election may be a more difficult course to follow.

This Element also focuses on the role of individual civil servants in opposing what they perceive to be illiberal or unconstitutional governments. To some extent, the decision to oppose the government and to take actions that are unconventional for public servants will be individual. However, there may be organizational and institutional factors that influence these decisions. For example, civil service systems may be more or less institutionalized in different political systems, and more institutionalized systems are likely to present less opposition to the policy demands of the government of the day. Also, some policy areas, and the organizations associated with them, may be more in the line of fire of a populist government than are others, and therefore may produce greater opposition, perhaps in order to survive,

Finally, we do not know what citizens would think of civil servants who choose to oppose the directives of their superiors in government, claiming that those orders were contrary to the constitution or to other laws. Some survey evidence, at least in the United States, reveals that citizens often trust civil servants more than they trust politicians, and therefore those citizens might be willing to accept a civil service that did resist pressures from their superiors in organizations (Hitlin and Shutava Reference Hitlin and Shutava2022; Partnership for Public Service 2022). Therefore, what may appear to some observers – perhaps especially the political leaders – as a usurpation of power, may be perceived as an appropriate response by the public servants, and as a means of supporting constitutional values.

6.2 Conclusion

This brief volume has addressed one of the most important issues of contemporary governance: how can civil servants committed to liberal democratic values work with elected governments that may not share those values? A closely linked question is how does the idea of an impartial, Weberian civil servant affect behaviors of those civil servants when fundamental constitutional and political values are being threatened. These have been, are, and will be, real questions in many democratic political systems.

The answers to the previously stated questions, when given by the civil servants themselves, are far from uniform or clear. Their behavior can be read in several ways. Several civil servants in the United States, Brazil and in other countries with illiberal leaders have chosen to leave office, but the vast majority have remained working for a government with which they may or may not agree. The evidence is that many have thought about leaving, but have chosen to stay, whether out of loyalty to the country or from personal desires to have a good career.

The answers to our questions in the survey experiment are also somewhat ambiguous. On the one hand, there were many more individuals willing to say they would shirk or sabotage than we might have expected, given the stereotypical conceptions of bureaucrats. On the other hand, there were also large numbers of civil servants who were willing to continue serving political leaders who had illiberal approaches to policy and governance. Those expressions of opinions might make advocates of the traditional civil service bargain happy, but they raise questions about the extent to which the civil service can be a bulwark against populist and illiberal values in government.

Our findings contribute to the growing literature on bureaucratic resistance, democratic backsliding, and the evolving nature of civil service neutrality. Traditional public administration theories, particularly Weberian bureaucracy, emphasize the neutrality and professionalism of civil servants, suggesting that their primary role is to implement policy directives impartially, regardless of their personal views (Weber Reference Weber1947, Reference Weber, Roth and Wittich1978). We already know that the separation between politics and administration did not produce that neutrality but, instead, insulated the bureaucracy to engage in its own politics (Peters Reference Peters2018). Our study further explores this more complex reality, where civil servants do not always act as passive implementers but instead navigate political pressures strategically, sometimes through acts of resistance.

Brehm and Gates’ (Reference Brehm and Gates1999) triad working, shirking, and sabotage motivated our theoretical and methodological approach. However, while evidence so far showed that bureaucrats most often prefer to work (Pierre and Peters Reference Pierre and Guy Peters2017), we demonstrate that democratic backsliding serves as a motivational trigger to turn shirking and sabotage into normatively desirable actions. This aligns with recent work on bureaucratic autonomy and resilience, which argues that civil servants are not merely reactive agents but can actively shape policy outcomes (Bersch and Fukuyama Reference Bersch and Fukuyama2023; Carpenter Reference Carpenter2002; Yesilkagit and van Thiel Reference Yesilkagit and van Thiel2008).

Additionally, our findings contribute to scholarship on democratic erosion and state capacity, particularly the debate on whether bureaucracies function as bulwarks against illiberalism or become co-opted by political leadership (Bauer et al. Reference Bauer, Guy Peters, Pierre, Yesilkagit and Becker2021; Bersch and Lotta Reference Bersch and Lotta2023; Yesilkagit et al. Reference Yesilkagit, Bauer, Guy Peters and Pierre2024). Building upon this literature, our findings show that resistance is conditional on economic security, hierarchical position, and tenure status, reinforcing recent arguments that bureaucrats resist selectively, based on both institutional constraints and personal risk calculations. This adds nuance to principal-agent models, which traditionally focus on mechanisms of bureaucratic control but often overlook how bureaucrats assess and manage risk in politicized environments (see, e.g., Gailmard and Patty Reference Gailmard and Patty2007).

Moreover, our research speaks to the literature on bureaucratic exit, voice, and loyalty, particularly in contexts of political instability (Guedes-Neto Reference Guedes-Neto2022; Hirschman Reference Hirschman1970; John Reference John2017). While some scholars argue that bureaucrats facing political interference must choose between loyalty (compliance), exit (resignation), or voice (active dissent), our findings suggest that shirking and sabotage represent a mix of active and passive dissent. These behaviors allow civil servants to remain in the system while still undermining illiberal policies, supporting recent arguments that bureaucratic resistance often takes indirect, covert forms rather than open confrontation.

By integrating insights from public administration, political science, and democratic theory, this study provides a more nuanced understanding of how bureaucrats navigate illiberal governance. Rather than viewing resistance as a binary choice between compliance and confrontation, our findings suggest that bureaucratic resistance is contingent, strategic, and shaped by both institutional structures and individual incentives. Future research should further explore the conditions under which bureaucratic resistance is successful or suppressed, as well as the long-term effects of bureaucratic noncompliance on democratic stability and governance capacity.

As global challenges to democratic norms intensify, the evolving relationship between civil servants and their political environments demands closer scrutiny. The willingness of some public servants to defy conventional expectations and resist illiberal policies underscores the importance of understanding the institutional and cultural factors that influence these decisions. Future research should delve into the mechanisms that enable civil servants to either resist or comply with illiberal pressures and the long-term implications of these choices for democratic governance. As the boundaries between administrative neutrality and political engagement continue to blur, the role of civil servants in upholding constitutional principles will remain a pivotal issue for both scholars and practitioners.

In sum, the ability of civil servants to defend democratic values will depend not only on individual commitment but also on the institutional frameworks that either support or undermine their autonomy. The questions raised by this study, though complex, are essential for understanding the future of democratic governance and the role that civil servants will play in preserving or challenging the norms that define it.