Fragmentation of Political Authority and Bureaucratic Entrepreneurship: Explaining Instances of Minority Accommodation in Israel and Estonia

Why do some ethnic nation-states committed to preferential treatment of the dominant nation choose to accommodate their ethnic minorities in some realms? I argue that power struggles between elected and non-elected officials account for the variation in the treatment of ethnic minorities. Fragmentation of authority creates opportunities for entrepreneurial bureaucrats to initiate policy changes and lead to unanticipated outcomes. Drawing on nationalism studies in comparative politics and principal-agent scholarship in public administration, this article outlines a theoretical framework focused on domestic factors accounting for variation in state policies toward minorities in a novel way. I apply this framework to education policy in Israel and Estonia vis-à-vis the Palestinian Arab and Russian-speaking minorities. This article illuminates an empirical puzzle of minority accommodation under nationalist governments and explains the conditions under which it occurs, offering generalizable theoretical expectations for similar contexts.


Introduction
Nationalizing states are defined as states of and for a particular nation, but at the same time as weak, unrealized, and insufficiently national in some sense (Brubaker, 1996). The elites that capture the nationalizing state use its apparatus and laws to address these perceived weaknesses to strengthen the status and rights of the dominant nation. This takes place in a context of ethnically heterogenous states, in which the ethnic minorities are excluded from the sociological boundaries of the state-owning nation defined in exclusive ethnic, cultural, or other terms. Discursive framing of these minorities as threatening to the survival of the dominant nation is used by governing elites to justify protectionist state policies.
Israel and Estonia are textbook examples of nationalizing states where state-owning majorities utilize the state apparatus and its laws to maintain their dominant status over the Palestinian Arab and Russian-speaking minorities, respectively. The State of Israel embedded Jewish ethnonational markers within the state definition, state symbols, and official ideology (Ghanem, 2001;Haklai, 2011;Peleg & Waxman, 2011). As a state of and for Jews (Kimmerling & Migdal, 2003), the Jewish population is the state-owning nation. The Palestinian Arab minority, comprising 21% of the population (Israel Central Bureau of Statistics, 2019), is excluded from the Jewish nation's ethnocultural boundaries and is marginalized by the state (Haklai, 2011).
Following its independence, the Republic of Estonia adopted an official discourse declaring ethnic Estonians as the legitimate owners of the state and the source of its sovereignty. The Estonian state became vested with the responsibility of protecting the Estonian nation in its putative homeland (Smith, 1996). The institutionalization of a dominant position of ethnic Estonians over the state led to rapid marginalization and exclusion of the Russian-speaking minority, which constituted about 30% of the population (Brubaker, 1996;Laitin, 1998;Schulze, 2010).
Alongside the ethnonationalist discursive frames and policies of exclusion, Israel and Estonia have also advanced some measures for minority accommodation. This is evidenced by the adoption of a dedicated state program for affirmative action in Israel and the preservation of the minority-language public education system in Estonia. Crucially for this paper, both instances of minority accommodation were adopted under nationalist governments committed to exclusionary ethnonationalist discourse on minority rights. This paper draws on nationalism studies in comparative politics and principal-agent scholarship in public administration to explain the unanticipated policies of minority accommodation in nationalizing states. The theory asserts that the extent of political authority's fragmentation over a given policy area influences policy outcomes toward ethnic minorities. State policies may deviate from the incumbents' preferred policy outcomes if political authority is fragmented. Under such structural conditions, non-elected officials leverage their expertise, networks and information asymmetry to redefine policy problems, build coalitions supporting policy innovations, and shift official government policies.
These findings contribute to the comparative scholarship on nationalism and ethnic diversity regulation. Power struggles and fragmentation of authority between elected officials over policy trajectories are central to the study of nation-building and state-minority relations. In this scholarship, policy decisions and their outcomes toward ethnic minorities are attributed to the political elites' preferences and motivations. Alternative explanations are relatively scarce. This paper seeks to shift our attention to another dimension of fragmentation of authority between elected and non-elected officials. There is a long tradition of analyzing this contestation and the role of bureaucracy in the policy process in public administration literature. Insights from this scholarship are analytically helpful for understanding the unanticipated effects of fragmentation of authority on policy outcomes in general, including those devised toward minorities. Building on this literature, this article problematizes the bureaucracy in nation-building processes and state-minority relations. To paraphrase Skocpol (1985), this paper seeks to "bring the bureaucracy back in" to nationalism studies.
Second, bureaucratic entrepreneurship in domains over which political authority is fragmented presents a pathway for accommodating other purportedly threatening minorities in majoritarian states. Migrant populations are one example. Contestation along preexisting identity-based lines, such as race, religion, or sexual orientation, is another one. That is to say that professional bureaucracies can present a counterweight to nationalist and populist governments.
The paper is structured as follows. In the first part, contemporary and classical research on policies toward ethnic minorities in ethnic nation-states is surveyed. The following section presents the theoretical framework of fragmentation of political authority. The remaining two sections outline the empirical analysis of two instances of minority accommodation in Israel and Estonia. The final section concludes the article and discusses policy implications.

State Regulation of Ethnic Diversity
In the comparative scholarships on nationalism and ethnicity, explanations of state policies regulating ethnic diversity are commonly rooted in domestic factors. Scholars have considered many factors, drawing on historical legacies, structural conditions, minority characteristics, and geography. The key developments in the scholarship on historical legacy and institutional design emerged out of the research on Central and Eastern European (CEE) countries due to the collapse of communist regimes (Beissinger, 2002;Frye, 2010;Laitin, 1998). In these states, the titular ethnic elites purposefully disadvantaged ethnic minorities to solidify ethnic ownership of the state. The logic of these explanations holds that the institutional legacy of the former political system and the status each group had held in it can explain the prospects for minority integration within them (Linz & Stepan, 1996). For example, the method of incorporation of the former Soviet republics and the treatment of the titular groups by the Soviet regime, ranging from restrictive in the Baltic states to relatively open in Ukraine, suggests a possible explanation as to why the former initially marginalized and discriminated against the Russianspeaking minority, whereas the latter accommodated this group (Laitin, 1998).
Historical legacy and institutional design are important explanations that have immensely contributed to our understanding of the structural and historical conditions for adopting state policies toward minorities. However, because these explanations focus on the initial selection of policies, they are insufficient to further our understanding of why policy change occurs over time (Aktürk, 2012). Furthermore, these explanations do not provide analytical tools for distinguishing between transitional and terminal nationbuilding policies, their implications for minority rights, and states' longterm intentions in developing these policies (Mylonas, 2015).
Ethnic conflict researchers have also offered explanations of state policies toward minorities based on past power relations between ethnic groups. A reversal in power relations between ethnic groups is expected to lead a previously marginalized, now-empowered group, to adopt exclusionary policies against the once-dominant ethnic group (Horowitz, 2000;Petersen, 2002). This expectation stems from the characterization of ethnic relations as inherently prone to conflict due to in-group favoritism and inter-group competition for group self-esteem and status (Horowitz, 2000). However, this theory, too, cannot explain policy change over time because it requires the occurrence of a structural change that removes institutional constraints supporting group power relations, thereby producing an opportunity for action (Petersen, 2002). Because the hierarchy between ethnic groups in nationalizing states is stable, this approach can only explain the initial policy selection but fails to explain policy changes occurring over time.
Other explanations for the development of state policies and their subsequent change use minority characteristics as predictors. States consider minority features of group size, territorial concentration, indigeneity, and transnationality when adopting policies (Buhaug et al., 2008;Cramsey & Wittenberg, 2016;Fearon & Laitin, 2003;Grigoryan, 2015;Horowitz, 2000;Posner, 2004). These characteristics offer parsimonious predictors for ethnic groups' behavior and the likelihood of mobilization against the state. Yet, because these explanations hinge upon relatively invariable characteristics of minorities-their size and territorial concentration-they offer little utility to understanding policy change and its direction.
Another strand of theories focuses on political elites' interest, agency, and mobilization strategies to explain the adoption or change of policies toward ethnic minorities (Cramsey & Wittenberg, 2016;Grigoryan, 2015). A significant contribution in this line was made by Aktürk (2012), who argued that a break from a previous policy regulating ethnic diversity would occur if new elites equipped with a new discourse on ethnicity come to power and succeed in institutionalizing their ethnic preferences. This and other elite-level explanations exemplify a common analytical problem of conflating elite preferences with their policy choices and the outcomes of these policies (Brubaker, 2011;Migdal, 2001;Mylonas, 2015). Migdal (2001) and Mylonas (2015) have accurately pointed out that the translation from one to another is at best imprecise. Elite-level explanations should not assume that elite intentions can be immediately observed through their policy choices or that policy outcomes are direct derivatives of elites' preferences.
Another common shortcoming of elite-level explanations is their reference to the ruling elites and nationalizing state as interchangeable, ignoring the fact that the state cannot be reduced to mean the governing political elites alone (Migdal, 2001). The appearance of state-owning elites as possessing powerful resources and the capacity to adopt and implement their intended policies should not be taken at face value (Haklai, 2007;Migdal, 2001;Mylonas, 2012). Indeed, elite preferences might be contingent on various factors, such as public opinion, institutional constraints, government changes, or economic crises, and, therefore, can suddenly change (Jenne, 2007). Haklai and Norwich (2016) have demonstrated the constraints imposed on political elites in Israel, often perceived as autonomous and powerful. Even when political elites were interested in including ethnic minorities in the governing coalition, inherited political traditions of minority exclusion fueled public opinion opposing such change, resulting in a missed opportunity for minority inclusion (Haklai & Norwich, 2016).
Overall, domestic-centered explanations implicitly or explicitly assume that ethnic nation-states states will either exclude or seek to eliminate diversity (Brubaker, 1996;Horowitz, 2000;Mann, 2005;Marx, 2002). This skew towards a general expectation of exclusion stems from the conceptualization of nationalizing states as coherent actors captured by majority elites, who utilize state apparatus to materialize the dominant nation's interests. This perception leads to the conclusion that the dominant elites would not act against their interests by adopting accommodative policies toward ethnic minorities. Consequently, an exclusionary policy better resonates with the dominant elite's preferences toward ethnic minorities.
These theories are insightful and important contributions offering parsimonious explanations for minority treatment in nationalizing states. However, their major conceptual drawback is undertheorizing why and when nationalizing states adopt policies of minority accommodation. While it is expected that nationalist elites will frame ethnic minorities as threatening in some way or another, discursive framing of policy preferences in nationalist terms does not necessarily translate into nationalist policy outcomes. Hence, in their current formulations, these theories have left the puzzling reality of accommodation under nationalist governments unresolved. Importantly, this occurrence is not rare or short-lived, as the empirical analysis in this article will seek to demonstrate. For this reason, we cannot dismiss this complex reality as being the result of an irrational leader or extraordinary circumstances.

Bureaucratic Autonomy
Policymaking in complex modern states is a joint venture of politicians and bureaucrats, where the latter play a central role in determining agendas and initiating policies, structuring policy priorities, and defining implementation strategies (Aberbach et al., 1981;Alesina & Tabellini, 2007;Gailmard & Patty, 2012;Page et al., 2005). Non-elected officials' policymaking authority stems from the asymmetry in professionalization, skills and expertise between them and the elected officials. This asymmetry can facilitate the reputational legitimacy government agencies enjoy and make them indispensable throughout the policymaking process (Carpenter, 2001;Carpenter & Krause, 2012). Moreover, reputational legitimacy affords them significant discretion "to enact outcomes different from the policies preferred by those who originally delegated power" (Epstein & O'Halloran, 1994, p. 699). Therefore, the ability to affect policy change is a form of fragmentation of political authority between elected and non-elected officials.
The extent of bureaucratic autonomy is partially determined by reputational legitimacy and the decision-making process in each policy area, reflecting historical legacies and governments' preferences to retain or delegate authority (Alesina & Tabellini, 2008;Carpenter, 2001;Skocpol & Finegold, 1982;Tsebelis, 1995). For example, governments typically choose to retain control over "hard" policies (such as foreign affairs and security) and policies with "unstable goals" (social policy) while delegating some authority on "soft" policies (such as environmental policy) and "stable" or highly technical ones (such as education and monetary policies) (Alesina & Tabellini, 2008;Hammond, 2003;Immergut, 1990). Political executives may also choose to retain greater control over essential policies for electoral and coalitionbuilding purposes but delegate policies that bring little political returns or expose them to electoral risks (Alesina & Tabellini, 2008).
Structural opportunities for increased bureaucratic autonomy may also appear due to the distribution of power among political actors following electoral cycles. In particular, the higher the number of fractions in the governing coalition and the more diverging their policy preferences, the more fragmented their authority; consequently, the more autonomous the bureaucracy can become (Hammond, 2003). Multiparty coalitions that lack consensus on a policy issue create structural opportunities for the bureaucrats to capitalize on and increase their autonomy from the elected officials in the policy process.
Bureaucratic entrepreneurship occurs when government agencies enjoy broad autonomy for policy innovation. Policies developed by bureaucratic entrepreneurs are distinguished from those adopted by agents external to the bureaucracy by their incremental evolution (Carpenter, 2001). Coalitionbuilding is also vital for these entrepreneurs' ability to persuade relevant political and interest-based groups of the value of their innovations and generate support for their adoption (Carpenter, 2001).
But which objectives do these entrepreneurs seek to achieve? Recent research on bureaucratic entrepreneurship helps theorize the types of policies bureaucrats in nationalizing states are likely to advance in "soft" policy areas, such as education. Civil servants in senior positions typically share the norms of evidence-based decision-making processes and a strong commitment to serving the public's needs (Andersen & Jakobsen, 2016). These values are most advanced among civil servants in social policies, including health care, education, crime control and corrections (Head, 2015). Moreover, socially aware behavior and orientation toward equity, not strictly efficiency in policy outcomes, is particularly evident in civil servants working in education, health, and social protection, especially among bureaucrats with higher education degrees in these areas (Fernández-Gutiérrez & Van de Walle, 2018). For example, evidence-based policy planning and the objectives of facilitating the economic growth of the society at large motivated Canadian bureaucrats to take a strong pro-immigration position between 1990 and 2010 (Paquet, 2015). They initiated this policy change without congruous political agenda, changes in societal perception on immigration, lobbyism or social movement mobilizing to this end (Paquet, 2015).
The utility of "fragmentation of political authority" is in accounting for agent preferences within a given institutional framework, both elected and non-elected. This allows us to analyze not only who gets to participate in the policy process and who does not, but also how agent preferences are weighed and which rules determine policy decisions (Immergut, 1990;Lowry, 2014). What theoretical expectations can we extrapolate about the type of policy a nationalizing state is likely to adopt toward its ethnic minority in each sphere and the likelihood of policy change if the political authority over that sphere is fragmented?
Given that the playing field in nationalizing states is skewed because of the domination of political actors representing the dominant majority, a policy of exclusion is the most likely outcome. This expectation is in complete agreement with the existing scholarship on nationalizing states. Based on this logic, a policy of exclusion is also expected to be highly stable as a result of relatively fixed preferences of majority political actors and their constituencies around minority treatment, the former's domination in the political arena, and electoral incentives to maintain a stable policy in areas under the direct oversight of the government. Consequently, I expect policy change away from exclusion if (1) moderate majority, minority-friendly, or minority-led parties manage to monopolize power in the governing coalition and (2) have electoral incentives to introduce accommodating measures. For example, few political parties with proximate ideological stances gaining political power can act as "minority advocates," introducing some measures of minority accommodation or moving away from exclusionary policies (Schulze, 2017). Under this scenario, a policy shift from exclusion to some degree of accommodation is possible.
I argue that this theoretical expectation will hold only when elected officials exclusively determine policies affecting ethnic minorities. But there is another overlooked path to policy change that does not necessitate political preferences to support a policy of minority accommodation. This path is possible under conditions of fragmentation when additional veto points exist along the decision-making chain, allowing other actors to alter the trajectory of the policy course. As described above, if non-elected actors, such as the military, the courts, the bureaucracy, or unions, are sufficiently authoritative, they can impose their policy preferences (Spruyt, 2005;Tsebelis, 1995). This expectation is supported by party politics scholars questioning the "parties do matter" hypothesis stipulating that policy outcomes reflect the ideological positions of political parties in the governing coalitions (Caul & Gray, 2000;Schmidt, 1996;Schnose, 2017). The assumption that public policies reflect party ideologies-such as that right-of-center governments minimize spending and produce lower deficits than left-of-center governments-is not empirically robust (Caul & Gray, 2000;Tavits & Letki, 2009). Research has found that political and economic environments can incentivize parties to act in opposition to their traditional platform stances (Tavits & Letki, 2009). Furthermore, the power of governments to determine policy outcomes is contingent on the extent to which these governments are "sovereign" in their decision-making process (Schmidt, 1996), or put differently, on the extent to which political authority is concentrated in the governing executive (Tsebelis, 1995).
Based on the research on the motivations of civil servants in the policy process, I theorize that meritocratically recruited bureaucrats are likely to seek to mitigate the impact of the central government's policies and even shift the status quo affecting the distribution of resources in "soft" policy areas and areas that allow high bureaucratic autonomy. Because professional bureaucrats are likely to follow an evidence-based policy approach, aim to improve service delivery, exhibit socially conscious behavior in social policies, and object to myopic policy planning, they are likely to oppose exclusionary policies towards a segment of the public they serve. These attributes can motivate bureaucrats to engage in entrepreneurial behavior in order to replace exclusionary policies with some extent of minority accommodation. That is because disenfranchised and marginalized minorities are likely to be less productive participants in the economy and in greater need of social and other state services. The benefits of adequate minority inclusion in the economy and society are likely to overshadow the suboptimal economic and social outcomes resulting from exclusionary policies.
To be sure, I do not claim that either individual or organizational motivations for policy innovation are driven by concerns for the welfare of ethnic minorities. Nor do I claim that bureaucratic entrepreneurs are motivated by an ideological commitment to champion minority rights. Some civil servants may even object to minority accommodation as a direct public policy objective. Instead, positive returns in the long-term on removing barriers excluding ethnic minorities from participation in the economy and the society can lead even ideologically nationalist bureaucrats to pursue plans for minority accommodation.
How well can these theoretical expectations explain the instances of minority accommodation toward the PAI in Israel and the Russian speakers in Estonia in the sphere of education? Before answering this question, the next section outlines the research strategy employed in this paper.

Research Strategy
This research employed a process-tracing analysis to develop causal mechanisms explaining how the fragmentation of authority leads to unanticipated policy outcomes of minority accommodation in nationalizing states. Process tracing is well-suited for identifying the central actors in the policymaking process and analyzing how they pursue their objectives within the institutional context in which they operate in a particular historical context (Spruyt, 2005). The analysis in both states focused on education-one of the most salient spheres for the nation-building process and the boundary-making between the state-owning titular groups and ethnic minorities.
This research draws on participant observations in roundtables, public discussions, and semi-structured interviews with 81 participants, including politicians, bureaucrats, and minority representatives. 1 The information gathered from interviews was used to learn about politicians' and bureaucrats' perceptions of the policymaking process and the institutional rules of "how things work." A secondary group of interviewees consisted of participants external to the policymaking circles but professionally engaged in the research or oversight over education policy pertaining to ethnic minorities. This group included representatives from minority civil society organizations, research institutions, and journalists. Interview data was cross-checked against official archival and census data, timelines of legislative procedures, and data from civil society organizations that monitor state policies. These sources allowed the tracking of policy development in each area, including changes in language, objectives, and the actors engaged in the policy process.

Case Selection
Israel and Estonia qualify as nationalizing states in relation to their transborder Palestinian Arab and Russian-speaking minorities, respectively. State policies toward these minorities exhibit a theoretically interesting variation, encompassing the typical outcome of minority exclusion expected by the scholarship and theoretically atypical or deviant outcome of minority accommodation. Israel and Estonia are also small states with substantial minorities ethnonationally belonging to regional majorities represented by the Russian and Arab kin-states. Due to the troubled relationships between the nationalizing and kin-states, ethnic minorities residing in the former are perceived as threatening and delegitimized groups. For these reasons, these countries with otherwise distinct historical legacy, language, religion, and regional environments make for compelling cases for comparative research (Smooha, 2002;Yiftachel & Ghanem, 2004).

The Education Policy of Israel
The education system in Israel is centralized under the authority of the Ministry of Education. It is composed of the general and religious education systems. Arab education was established as a separate stream within the general education system, yet unlike the religious schooling system, it was not granted autonomous status (Shafir & Peled, 2002). Israeli government utilized this separation under a centralized authority of the Ministry to target the minority education stream with discriminatory policies. The minority was excluded from the decision-making processes on resource allocation in the Arab stream, which allowed the Ministry of Education to underfund and disqualify Arab schools from state subsidies (Al Haj, 1995). Similarly, Arab educators were banned from the decision-making process on the curriculum and therefore prevented from them the possibility of determining the educational content of their national group (Jabareen & Agbaria, 2010). As a result, for over four decades, the curriculum in Arab education lacked references to the Palestinian Arab history, national identity, literature and culture.
On December 30, Israel's government, led by Benjamin Netanyahu from the Likud party, adopted a resolution on "Government Activities for Economic Development in Minority populations, 2016-2020," also referred to as Government Resolution 922. The Resolution was a comprehensive program for state investment to develop the Palestinian Arabs' human capital and increase the minority's integration in the national economy (Ministry for Social Equality, 2017). The government approved a budget of 15 billion NIS to implement this plan, which encompassed the spheres of education, industry, infrastructure, employment, trade, housing, and culture (Haj-Yahya & Asaf, 2017). The comprehensive plan and the substantial budget for its implementation make this the largest and highest-funded state program ever invested in the PAI (Haj-Yahya & Rudnitsky, 2018).
Of the total budget of 15 Billion NIS, almost 6.8 Billion NIS (approximately 40%) was allocated to education. This budget share made education the principal domain of state investment in the PAI. Some 915 Million NIS were dedicated to informal learning, including summer camps, youth enrichment programs, community centers, and improving the teaching quality in Arab schools (Kasir & Tsachor-Shai, 2016). This investment in informal education in PAI localities marked the first allocation of state resources of its kind (Haj-Yahya & Rudnitsky, 2018).
In addition to the increased resource allocation, Government Resolution 922 also marked an important change in the decision-making process relating to Arab education. Historically excluded from the decision-making process, PAI representatives were included in policy formulation and implementation processes (Shinwell et al., 2015). The Ministry also expanded the Unit for Arab Education from three to 15 civil servants (Haj-Yahya & Rudnitsky, 2018). The Ministry also sought public input on informal education activities that would meet the needs of Arab children and youth. Over 1500 respondents participated in this public consultation process (Haj-Yahya & Rudnitsky, 2018).
This accommodative policy is an unanticipated outcome for the PAI as it was adopted at a time of the domination of nationalist governments. Rightwing governments led by the Likud party (2001-2006; 2009-2019) maintained a nationalist discourse and actively discriminated against the PAI. In 2011, a Likud-led government introduced a ban to commemorate the Nakba (literally translating to "catastrophe," which is how Palestinians refer to the establishment of Israel) in all education programs and intensified the Zionist historical accounts in the curricula (Jabareen & Agbaria, 2010). Similarly, during the electoral campaign of 2015, targeting the PAI, at least rhetorically, was part of Netanyahu's strategy for mobilizing his electoral base. He notoriously urged his political base to turn out to vote because the "Arabs were flocking to the voting booths in buses driven by left-wing NGOs" (Kaspit, 2018). Once in power, this Likud-led government advanced a nationalizing agenda and produced a series of discriminatory laws against the PAI. Among them were the NGO Transparency Law from 2016, which was motivated politically to target human rights NGOs, and the controversial Basic Law: Israel Nation-State of the Jewish People from 2018. The latter defined Israel as the state of the Jewish people only, thereby restricting the right of national selfdetermination within Israel to Jews alone (Adalah, 2018). It also declared Hebrew as the (only) official language in Israel, relegating the Arabic language from an official language to one with an unspecified special status.
Overall, the government elected in 2015, the year the Resolution was adopted, was a Likud-led coalition of right-of-center and religious parties committed to a nationalizing imperative. While the former had a general electoral appeal to the middle and lower socioeconomic classes, the religious parties represented niche constituencies with particularistic interests, specifically, the orthodox Jews, settlers in the West Bank, and Mizrahi Jews (Jews of North African and Middle Eastern descent). Accordingly, none of these parties represented the PAI. In fact, they potentially stood to be punished electorally for promoting the interests of the PAI minority at the expense of those of their constituencies.

Explaining Israel's Policy of Minority Accommodation
What explains the divergence between this government's level of commitment to a nationalist agenda and the accommodative program? The answer lies in the fact that the Likud-led government did not initiate this program. Instead, the absence of an explicit agenda toward the PAI (other than the general preference for their discrimination) created a structural opportunity for bureaucratic drift. Comprehensive bureaucratic reforms during the preceding decade provided the motivation for bureaucratic entrepreneurship to shape policies toward the PAI. Compelling definition of minority exclusion as a policy problem, framing of the Resolution as a strategic plan for national economic development, and collaborative storytelling and coalition-building for its support persuaded this right-wing government to defer to the expert authority of the bureaucrats. The following section details the processes within the bureaucracy that led to the adoption of Resolution 922.

Developments in Israel's Civil Service
During the first decades of Israel's independence, the political system was dominated by the hegemonic Labor party, in power between 1948 and 1977. Through integration and cooptation of pre-state institutions, and the establishment of new ones, Labor effectively controlled the formation of the public administration and created strong links between the party and the bureaucracy (Nachmias & Sened, 2002). The heavy politicization of civil service had meant that these officials were not meritocratically recruited and promoted, but rather that they worked at the party's service (Etzioni-Halevy, 2002). As a result, the bureaucracy lacked impartial and formal rules of conduct, was insufficiently professional, and overall characterized by favoritism so pervasive as to become an accepted institutional practice (Arian, 2005;Nachmias & Sened, 2002;Rosenbloom & Yaroni, 2001). Even with the end of Labor's hegemonic rule in 1977, Israeli bureaucracy did not become more professional and independent of the governing political elites because subsequent governments avoided the global wave of public administration reforms of the 1980s and 1990s (Cohen, 2016).
It was not until the late 2000s that the civil service in Israel underwent a sweeping reform targeting its efficiency in policy planning, holistic policy development and transparency in service delivery and policy implementation. In 2011, Israel's public administration underwent additional reforms due to the cumulative effects of joining the OECD in 2010, recommendations of special commissions to improve the management of public administration, and public protests that demanded responsive and costeffective public services (Cohen, 2016). These reforms encouraged and rewarded bureaucratic entrepreneurship, which in turn, had significant implications for state policies toward the PAI. The development of what would materialize into Government Resolution 922 illustrates this.
In 2013, the National Economic Council, the Ministry of Education, and the Authority for the Economic Development of the Minority Sectors developed a strategic policy for Israel's economic development. According to their evaluation, Israel's economic advancement was entering a stagnation phase, and because it depended primarily on human resources, the reversal of the anticipated stagnation trajectory required maximizing all Israeli residents' human potential (Shinwell et al., 2015). The writers argued that the discrimination against the PAI in various life areas marginalized them from the economy and constituted the key impediment to the strategic economic advancement of Israel as a whole (Shinwell et al., 2015). Accordingly, they recommended a significant government investment in the PAI. Education was identified as the most influential factor for employment opportunities and a key area for state investment (National Economic Council, 2015).
This committee's analysis and recommendations gained traction among the senior and strategically positioned civil servants, who concurred with the idea that "What is good for Arabs, is good for everyone" (IS28, Ayman Odeh, Knesset Member, personal communication, May 17, 2016). The program's initiators-the National Economic Council, the Ministry of Education, and the Authority for the Economic Development of the Minority Sectors-were joined by new partners in government ministries, including the Ministry for Social Equality and the Ministry of Finances.
Strategically positioned bureaucrats staked their reputation on this program. A notable example of an entrepreneur behind this program is Mordechai Cohen, a senior official at the Ministry of the Interior at the time. As one of the key initiators of the program, Cohen capitalized on his reputation as an influential policy entrepreneur known for his commitment to reducing socioeconomic inequalities (Arlozorov, 2018). Another highly influential official behind the program was Amir Levi, the head of the budgets office at the Treasury. In 2014, Levi met with dozens of heads of Arab local authorities and urged them to "take responsibility for their destiny" and actively lobby the state and the bureaucracy for budgets (Barkat, 2019). He explicitly encouraged PAI representatives to develop close ties with treasury officials and senior bureaucrats across all ministries directly responsible for managing PAI localities' budgets. Levi also insisted on the involvement of the heads of the Arab local authorities in developing the Resolution (Barkat, 2019), a practice that broke with the previous norm of minority exclusion from the decisionmaking process.
Scholars have demonstrated this ability of Israeli bureaucrats to set political agenda and lead to policy changes (Nachmias & Sened, 2002;Pedahzur, 2012;Spruyt, 2005). One conspicuous example is Cohen's (2012) study of the successful entrepreneurial behavior of non-elected officials under the Rabinled government that led to Israel's National Health Insurance Law in 1994. The policy was driven by influential bureaucratic entrepreneurs that recognized structural opportunities for policy change, employed strategies to set the policy initiative on the political agenda, and built coalitions in its support (Cohen, 2012).
Another successful way of influencing the political agenda is attributable to the Israeli bureaucracy's close relations with organized sectarian and interest groups. According to some, the characteristically close ties between Israeli ministries and organized interest groups can be politically beneficial to both sides (Nachmias & Sened, 2002). In return for the advancement of their policy preferences, interest groups make their political resources available to bureaucrats in government ministries seeking to increase their capacity to influence public policy, including lobbying political representatives on the bureaucracy's behalf, engaging in public relations regarding the Ministry's activities, and assisting in their implementation (Nachmias & Sened, 2002). Similarly, the high representation of organized interests within the Israeli bureaucracy has been conducive to these clients' ability to successfully set a policy issue on the agenda and affect policy change. For example, organized agricultural interests represented in the water policymaking authorities in Israel successfully defined water policy problems and shaped their solutions (Menahem, 2001). They even framed national interests in water to benefit agricultural interests, despite the irreversible damage of overexploitation of limited water resources such a policy agenda eventually caused (Menahem, 2001).
These examples illustrate the agenda-setting capacity of the Israeli bureaucrats in various policy spheres. However, the adoption of new policy initiatives requires political approval, or at the least, no outright political objection. This was no exception for the bureaucratic initiative to advance the national program for minority accommodation, which had to be approved by a government highly committed to the "nationalizing state agenda." Typically, this awareness of the need to secure political support leads senior bureaucrats to plan policies based on the reactions they anticipate from the political actors and seek to address potential sources of conflict before they materialize (Aberbach et al., 1981;Carpenter & Krause, 2015;Page et al., 2005). A senior Jewish official at the Ministry of Economy explicitly expressed an anticipatory behavior of civil servants who promoted this agenda before its presentation to the government: "When coming to the political level, we need to have a story...and we had many partners to this storytelling, such as the National Economic Council, the Bank of Israel and others. And all of them engaged in this storytelling to the [political] policymakers" (IS21, Senior official at the Ministry of Economy, personal communication, March 29, 2016).
Civil servants framed policy initiatives for comprehensive minority accommodation as a strategic plan for benefitting the national economy (National Economic Council, 2015;Shinwell et al., 2015), not as an ideological commitment to minority accommodation. This policy framing was a fundamental objective for Netanyahu (and indeed most politicians), and something he has eagerly credited to his own efforts (Bahar & Eckstein, 2019).
That professional bureaucrats rather than elected officials developed the Resolution provided it with the advantage of a comprehensive implementation plan which was highly detailed with an annual allocation of budget and expected implementation rates (Haj-Yahya & Asaf, 2017). These features of a gradually evolving and well-specified implementation plan distinguish policies originating in the bureaucracy from the ones promoted by politicians (Carpenter, 2001).
Two years after Government Resolution 922 was adopted, senior officials continued to present a unified narrative on the necessity to accommodate the Arabs for Israel's strategic economic development. Cohen, one of the program's initiators, stated in 2017: "The Israeli government is not doing the Arab society any favours…[accommodation] does not come from a humane or values-based place…it is clearly in the government's interest, and if it [accommodation] does not materialize, the state will have given up a huge potential" (Lavie, 2018). David Brodet and Eli Gruner, the Chairman of the Bank Leumi Board of Directors and Director-General of the Prime Minister's Office, respectively, articulated similar ideas and framed minority accommodation as a necessary step development of Israel's economy (Lavie, 2018).

The Education Policy of Estonia
Following independence, the first post-Soviet government authorized the formation of a new education system. Through legislative and administrative steps, Estonia's education policy was set out to achieve two objectives: creating an independent, decentralized education system and de-Sovietizing its content and goals.
First, in 1992, Estonia adopted educational reforms modeling decentralization and school autonomy objectives after the Finnish educational system (Byrne & Plekhanov, 2019). These reforms manifested in the institutionalization of complementary responsibilities and jurisdictions between the Ministry of Education, local governments, and the schools. Local authorities oversaw and managed primary education, while the Ministry of Education managed upper-secondary schools. The schools exercised a significant degree of autonomy in managing school affairs, curricula, educational goals, hiring teaching personnel, and allocating school budgets (Ministry of Education, 2001). As a result of these complementary responsibilities but diverging spheres of autonomy, negotiations between these bodies became an integral feature of the decentralized education system (OECD, 2001).
Notably, the Education Act required all secondary schools (grades 10-12) to transition to Estonian as the primary language of instruction by 2000. With the introduction of the University Act from 1995, the option to pursue higher education in Russian in a public university in Estonia was eliminated. These reforms foreshadowed the decline in the former dominant status that the Russian language enjoyed in the education system.
De-Sovietization was the second objective of the education policy. A new national curriculum was adopted in 1996, focusing on new learning materials and pedagogical retraining of teachers trained during the Soviet era (Lees, 2016). The reformed system adopted a Western-oriented pedagogical approach and phased out the Soviet one. This transition entailed transforming the learning process from a strictly hierarchical teacher-student relationship to more pluralist and student-focused interactions. Learning outcomes prioritized the humanities and active learning rather than hard sciences and memorization of material (Ministry of Education, 2001).
The structural and substantive changes to the education system were profound, as were these changes' practical implications. The decentralized structure and differential sources of authority over schools allowed the Russian-speaking schools to exercise autonomy in defining educational objectives, choosing learning materials, the language of instruction, and pedagogical approaches, thus creating significant disparities between Estonian and Russian-speaking schools (Krull & Trasberg, 2006;Lees, 2016). These schools established Russian as the dominant language of instruction, and between 1992 and 1996, they taught Estonian as a foreign language, introduced gradually only in the third grade. Further, Russian language schools followed Soviet-era curricula and continued to use learning materials produced in the USSR and later in Russia (Vetik, 2002). They also maintained the priority of subjects in a way characteristic of the USSR; favored hard sciences over humanities; and employed a centralized, top-down hierarchy within schools-the Soviet era's pedagogical approach that was abolished in Estonian language schools (Krull & Trasberg, 2006;Vetik, 2002).
That this policy was adopted by the first government committed to a nationalizing agenda is an unanticipated outcome. This right-wing coalition came to power with the promise to enact laws and government programs to protect ethnic Estonians and "clean the house" of Russian speakers (Pettai & Kreuzer, 1999). Ethnicity was central to this right-wing government's agenda and inherently linked to questions of national identity and survival of the Estonian culture and language (Lauristin & Vihalemm, 2009;Vetik, 2019). The Pro-Patria-led government perpetuated a discourse of the existential threat the Russian speakers posed to the survival of the Estonian nation and sovereignty, framing them as occupants or tools of the Soviet occupation (Commercio, 2008). The minority, which constituted over one-third of the population, was said to be too large to be assimilated (Laitin, 1998). Ethnic grievances toward Russians were at the heart of policies designed to reverse the previous hierarchy between the groups and ascertain Estonian domination (Vetik, 2019). In education, this was demonstrated by the rapid unfolding of reforms in Estonian schools, which among other things, sought to change the previous asymmetric bilingualism favoring Russians. The neglect to reform minority schools on par with Estonian schools created long-term educational disparities favoring Estonian and structurally disadvantaging and marginalizing Russian speakers (Byrne & Plekhanov, 2019;Vetik, 2019).
And yet, this government did not abolish the Soviet-era separate education stream for Russian speakers. Neither did it enforce tight control over it, the way Israel did for over half a century. Instead, the decentralized education system institutionalized partial autonomy for Russian speakers in education. This autonomy undermined the government's attempts to institutionalize a nationalizing system in which the minority would be stripped of all its past privileges. Russian speakers, in turn, strongly objected reforms aimed at increasing Estonian language teaching, viewing them as assimilative measures (Vetik, 2019). The separate education streams meant that the Russian schools were not fully accommodated as state education facilities that provided education to a distinct community, yet they were not expected to be fully integrated into Estonian society, of which they became an integral and significant part (Agarin, 2010).

Explaining Estonia's Policy of Partial Minority Accommodation
Partial minority accommodation of Russian language schools adopted by Estonia's first independent government resulted from the fragmentation of political authority over education policy between politicians and bureaucrats. The following section discusses how and why bureaucrats were able to use their leverages over the government to reform the education system and institutionalize partial minority accommodation.

Developments in Estonia's Civil Service
The Estonian educators and education officials' broad policymaking authority stemmed from a pre-independence organizational structure, their role in the independence movement, and their capacity to set policy agenda. Their capacity for collective action had begun to form already in the late 1970s as a response to the aggressive Russification campaign undertaken by the central authorities in Moscow. Estonian educators' strong resistance to this reform succeeded in postponing its implementation and even made Estonia one of the last Soviet republics to align with this Soviet-wide policy (Grenoble, 2003). At the Teachers' Conference in 1987, Estonian educators and education officials voiced deep grievances over the marginalization of the Estonian language and national content in the curriculum and the prevalence of Soviet educational values that prioritized memorization over the practical application of theoretical knowledge (Krull & Trasberg, 2006). Their mobilization against Russification gained broad traction during the national independence movement, ultimately leading to the passage of the Estonian Language Law in 1988, which contravened the all-Soviet language policy of Russian language domination in the public sphere (Krull & Trasberg, 2006).
Between 1989 and 1992, forums on education revival facilitated by Sovietera education institutions of the Forum of Culture and Education and the Council of Education drew large participation from both professionals and the public (Ministry of Education, 2001). These bodies were the precursors of the Estonian Education Forum, laying the foundations for redesigning Estonia's national education system (Krull & Trasberg, 2006;Ministry of Education, 2001). This cooperation among Estonian educators resulted in a "…strong, indigenous, grassroots movement for educational renewal" that laid the institutional foundation for the Estonian education system (Ministry of Education, 2001, p. 8).
By state independence in 1992, education officials already relied on a robust institutional backbone and a set strategy for educational reform. The organizational strength and agency of education officials, coupled with the inherent weakness of the first government in power, had meant that the political authority over education was fragmented. Education officials capitalized on their expertise and authority to define the centralization and politicization of education as grave problems associated with the experiences of Soviet control that needed to be urgently solved (Byrne & Plekhanov, 2019;Sarapuu, 2016). They enjoyed broad authority in strategic planning and policy leadership (OECD, 2001), making them able to exercise "…considerable leverage over the issues belonging to their areas of governance" (Uudelepp et al., 2013, p. 4). Education officials successfully advocated for the adoption of a Western-oriented approach of decentralized decision-making in education to allow local authorities and schools autonomy to define their goals and express the interests of parents and students (Ministry of Education, 2001).
Further autonomy was later accomplished by institutionalizing the division of responsibilities between state actors in the policy process. Government ministries are primarily in charge of policymaking, while policy implementation is carried out by state agencies established for this purpose (Uudelepp et al., 2013). In education, the Ministry acts as a policymaker, state foundations such as Innove and MISA act as implementing bodies of state policy, and the think tank PRAXIS, Tartu and Tallinn Universities, and Narva College, act as policymaking partners (EE26, Senior official at the Ministry of Education, personal communication, July 11, 2016). Although these bodies are responsible for policy delivery, their influence on policy can also be high due to the concentration of expert knowledge and limited resources of ministries. Thus, even subordinate agencies vested with implementation can influence policy as well as the regulations that govern their work (Sarapuu, 2016).
The leverage education bureaucrats had over education policy is also evident in the depoliticization of the education system, a key priority of the reform. Education Ministry officials partnered with the Academic Council of the President of the Estonian Republic (who is a ceremonial figure) to develop a national strategy "divorced from party politics" (Krull & Trasberg, 2006). To limit the influence of incumbent governments on the national curriculum, the Ministry delegated authority for its development to an external professional body of the Laboratory of Curriculum Studies at the Tallinn Pedagogical University (Krull & Trasberg, 2006).
According to education officials, the influence of the ministers on policy has been rather limited. Their tenure is typically associated with specific programs, such as digital development or promotion of science, which are discontinued after a ministerial turnover (EE26, Senior official at the Ministry of Education, personal communication, July 11, 2016). This assertion aligns with survey findings showing that Estonian civil servants' self-perception as decision-makers is relatively high. Only 22.2% claimed that politicians, not senior civil servants, initiate new government reforms and policies; nearly 40% disagreed with this statement (Savi & Metsma, 2013).
Defining centralization as a policy problem, seeking to overhaul the education system rather than devise gradual and specific solutions, and building strong networks-based support for educational reform within the preindependence movement are distinctive features of entrepreneurial bureaucratic behavior. Although decentralization was intended to secure ethnic Estonians' rights and freedoms, the proximate unintended consequence was the extension of protections and autonomy to the Russian minority. Accordingly, the abolition of Russian schools was not feasible given the education system's decentralization and autonomy at the local level (Agarin, 2010). Consequently, fragmentation of authority undermined the government's attempts to enact a nationalist agenda in education, leading to partial accommodation of this minority, even though it was not the intended objective of neither the education officials nor the politicians. Estonian officials and educators' commitment to a decentralized and depoliticized education system that affords schools autonomy was repeatedly conveyed by education officials interviewed in this research (EE26, Senior official at the Ministry of Education, personal communication, July 11, 2016; EE33, Senior officials at the Ministry of Education, personal communication, July 25, 2016; EE19, Senior officials at the Ministry of Education, personal communication, June 16, 2016). A senior official at the Ministry of Education explained the importance of decentralization and school autonomy, as well as their immediate implications for minority-language education, in the following way: "To achieve equality in education, diverse needs of students in different schools and regions must be taken into account…Only teachers know what the students in their schools need. If the teacher is not free to decide which tools are best for their students, then we cannot achieve equality. If a person is not free to decide, he will never take responsibility. If the teachers and headmasters are not free to choose the methods, then they are not ready to be decision-makers" (EE26, Senior official at the Ministry of Education, personal communication, July 11, 2016).

Conclusion
This article examined the conditions under which nationalizing states committed to the preferential treatment of the dominant nation at times adopt policies of minority accommodation. Comparative scholarship on nationalism and ethnicity has offered various theories on the adoption and change of state policies toward minorities, most commonly emphasizing historical legacies, past power relations between ethnic groups and ethnic conflicts. Scholars have generally examined minority policies as direct outcomes of dominant political elites' perceptions and policy preferences toward these groups. Given the uneven playing field between ethnic groups in nationalizing states, inclusionary policies are likely to occur where the minority enjoys institutional guarantees protecting its rights or political representation in government. Findings presented in this research suggest that minority inclusion also occurs in the absence of minority-dominant or minority-friendly parties in the government or institutional guarantees for minority rights.
This research highlights an understated feature of policymaking in nationalism and ethnic studies scholarships: the agency of non-elected officials in influencing policy outcomes. Minority representation in the state bureaucracy, where some policy trajectories are set, and most practical measures are decided upon, is central to determining state policies toward these groups. Still, in nationalism studies, it has not been subject to rigorous comparative empirical analysis, nor has its effects on policy outcomes been well specified.
My theory suggests that the extent of authority fragmentation over a policy issue influences policy outcomes toward ethnic minorities. Elected officials adopt policies in accordance with their political preferences and have the oversight capacity to determine desired policy outcomes if they concentrate political authority in the executive branch. Thus, nationalist governments are likely to adopt exclusionary policies, and minority-friendly ones are likely to adopt more accommodating ones. The concentration of political authority over the policy process increases policy stability, making the policy change contingent on the incumbents' alternation in power.
However, where political authority is fragmented, additional actors become endowed with policymaking authority. Drawing on process-tracing analysis and interviews with elected and non-elected officials, I show that fragmentation of authority creates opportunities for entrepreneurial behavior seeking to innovate on existing policies and cause diversion from the government's policy preferences. Bureaucratic entrepreneurship can explain why, at times, even nationalizing states adopt policies toward minorities that are inconsistent with the official government preferences.
The Israeli case illustrates how the fragmentation of authority over education policy enabled bureaucrats to advance minority accommodation under a government with a strong nationalist agenda. The non-elected officials capitalized on the absence of a clear government policy toward the PA to introduce policy innovations. Evidence-based policy planning led these officials to define minority exclusion as detrimental to the growth of the national economy and initiate a policy shift toward minority inclusion.
In Estonia, education officials enjoyed autonomy due to the solid preindependence structure and high fragmentation of political authority. Resentment of Soviet-era centralization and politicization motivated them to act against dominant elites' attempts to enact similar measures in independent Estonia. Education officials used their leverages to institutionalize a decentralized education system, resulting in a more accommodating outcome than the governing coalition at the time had preferred.
Two alternative explanations focused on external actors' influence on nationalizing states' policies deserve a brief discussion. First, regional organizations can compel nationalizing states to steer away from radical solutions to ethnic diversity and provide incentives to accommodate them. Accordingly, some scholars attribute the positive changes in minority rights in CEE states to pressures and incentives European organizations employed as conditions toward prospective member states in the European Union (Kelley, 2004;Vachudova, 2005). In Estonia, European actors scrutinized legislative and political processes in the late 1990s and closely monitored the situation of the Russian-speaking minority before Estonia acceded to the EU in 2004 (Agarin & Regelmann, 2012).
Second, ethnic kin-states have also been argued to have played a role in certain CEE states' domestic policies (Brubaker, 1996;Schulze, 2017). Indeed, Russia, the homeland of Russian speakers in Estonia, used various economic, political and military methods to influence Estonia's treatment of its kin group. Similarly, the transborder Palestinian Arab minority shares cultural, religious, linguistic and other ties with Arab states surrounding Israel. Here too, the troubled relationship between the kin and nationalizing state, with a history of conflicts, presents the backdrop against which domestic actors develop minority policies.
Ultimately, these explanations fail to account for the variation in state policies in both cases due to the lack of consistency between the timing of external pressures the domestic elites experienced and their decisions to change policy toward the minority. Estonia accommodated the minority in education already in the early 1990s, some years before it signaled its interest to join the EU, and restricted it in the 2000s, after EU accession in 2004. It also adopted its initial policy before the peak of Russia's pressure in the mid-1990s and restricted access to minority-language education despite renewed Russian pressures in the mid-2000s. In other words, the timing of external pressure does not correspond to the timing and direction of the policy formation period. There are no similar regional organizations in the Israeli case to speak of. While the Arab states surrounding Israel have sought to influence Israel's policies, these have focused on Israel's treatment of the Palestinians in occupied territories and the Palestinian Authority, not Palestinian Arab citizens of Israel. Thus, key decisions and policy choices on minority issues remain domestically driven by nationalizing states (Agarin & Regelmann, 2012).
The implications of these findings raise important questions for political science researchers who examine state regulation of ethnic diversity. The argument that fragmentation of political authority may have unanticipated effects on state policies departs from the assumption in nationalism studies that ethnic nation-states are coherent actors that develop well-structured and purposeful policies. This analysis puts under the spotlight the engagement of bureaucrats with politically contentious issues, a dynamic presently overlooked in nationalism scholarship. It also highlights the structural counterpoint that professional bureaucracies can present against nationalist and populist governments. In the untidy give-and-take political process, different domestic actors negotiate and compromise, form alliances and instigate political crises, and routinely engage in entrepreneurial or risk-averse behavior. Thus, we need to further our understanding of the ways in which different domestic actors' access to networks, resources, skills, and information shapes the state treatment of minorities.
This argument also has broader applicability beyond the study of nationalism and ethnicity. Woodrow Wilson famously viewed the civil service as an apolitical institution of public trust that exists to carry out state policies. He said: "Administration lies outside the proper sphere of politics. Administrative questions are not political questions" (Wilson, 1887, p. 210). This paper presented evidence that defies Wilson's position. It also gives reason to believe that professional bureaucracies may seek to improve the well-being of the public they serve, which is far greater than the public majoritarian governments represent. This suggests that the treatment of minority groups is strongly intertwined with the consolidation of professional bureaucracies that are not beholden to the incumbent governments. Consolidation of such bureaucracies is likely to facilitate more equitable treatment of disadvantaged groups. Of course, minority-friendly governments would make the likelihood of minority inclusion even greater.