A policy cycle of electoral reform: comparing directly elected chief executives in Israel and Turkey

In studying electoral reforms, the democratic context is often a neglected variable. Israel and Turkey, at different times with different levels of democracy, have implemented a similar reform, the direct election of the chief executive. While there were similar justifications at the time of the introduction, the reforms then took on different dynamics. In Israel the reform was reversed, while in Turkey it became the centrepiece of further autocratisation. Adapting a framework of policy-analysis, the article explores similarities and differences in both cases. It argues that the democratic environment turns out as an important factor to explain the different trajectories of seemingly similar policy reforms in both cases. While the reverse reform in the Israeli case is linked to advantages in processing expert and civil society information, the trajectory in the Turkish case is based on the different power structure in the course of presidentialisation. Applying a comparative policy analysis to these cases of electoral reform, the article concludes that the democratic environment, that is, the degree of freely available information, helps explaining the different trajectories in both cases.


Introduction
Electoral reform is not among the typical fields of (comparative) policy analysis. Yet the policy-cycle model offers a structured framework, especially for small-N comparison and the timing of changes. It is therefore particularly suitable for a structured analysis or comparison of the process dimension of (electoral) reform policies (Kortukov 2020). Policy models sensitize for the tension between power interests and factual considerations 1 as well as sector-specific factors (Knill and Tosun 2020). For instance, in electoral reform the "epistemic" or policy community is rather marginal, and the balance between power interests and factual considerations can be considered leaning towards the former. With Theodore Lowi's typology, electoral reform can be characterized as redistributive policy, which is usually particularly conflictual due to its polarization of winners and losers. A reform of electoral law entails a redistribution of seats and thus of power. Even if the process often resembles a muddling-through, for the actors involved it is also a matter of "who gets what, when, how" (Lasswell 1936). This power interest can be controlled or counterbalanced by a vigilant and critical public and by policy experts. However, this presupposes conditions such as a democratic, participatory public sphere. One can therefore assume that inter-elite negotiation processes and elite-mass interaction (Renwick 2010) also depend on the degree of democratization, which, however, is often neglected in rational choice and elite-centred studies on electoral reform (Boix 1999;Bowler, Donovan, and Karp 2006).
The cases of Israel and Turkey have been chosen as they introduced similar reforms (direct election of the chief executive) albeit at different times. Both political systems have been diagnosed with forms of "governability crisis" (Rosenthal 2016;Kenig 2021;Heper 2011) at the time the reforms were praised as a panacea against unstable coalition governments, gridlock and snap elections. At closer inspection, however, already the problem definition shows significant differences, which become even more visible as the policy cycle progresses. While in Israel the system of checks and balances remained largely resilient against a power grab by a populist prime minister, in Turkey a populist strongman has changed the institutional setting like no politician before, except Ataturk (Cagaptay 2020;Christofis 2018). In Israel, the experiment with a directly elected prime minister began in 1992 after a time of discontent in the early 1990s with the formation and performance of coalition governments, and was rescinded about a decade later. In Turkey the direct election of the president, introduced by law in 2012 and first applied in 2014, turned out as a core element of further autocratisation, despite the cooperation of meanwhile six opposition parties for a return to a strengthened parliamentary system (Küçük 2021).
Against this backdrop, the article asks why seemingly similar reforms gave rise to so different developments. In other words, what makes the Turkish case so "susceptible" to such an executive strengthening reform, compared to Israel? It is argued that the often neglected democratic context influenced the policy cycle from problem definition to the feedback loop via the plurality of information available to the public. The article aims to focus attention at the level of democracy as a factor influencing the course of electoral reforms. Section two introduces the policy cycle and the idea of a directly elected chief executive. Section three applies this heuristic structure to the case material, while section four concludes.

Theoretical framework and proceeding 2.1 Theories of electoral reform
Rational choice and elite centred approaches, often with a strong focus on intra elite negotiations and compromises, less often on mass-elite interaction (Renwick 2010), are prevalent theoretical approaches to electoral reform. This perspective seems particularly suitable for explaining the benefit calculations behind reform initiatives and the relative stability of electoral systems (Rahat 2011). It analyses reform initiatives and strategies as derived from the goal of interest maximisation. For instance, it has been found that candidates of both ruling and opposition parties who emerge as winners of an election have a higher satisfaction with the electoral status quo and thus a lower incentive for reform (Bowler, Donovan, and Karp 2006).
Both the rational choice and the elite approach are status quo centred. This raises the question of why there are still so many (or few) reforms "from above" (Katz 2008). The cartel party theory (Katz and Mair 1995) explains this with foreclosure tendencies of established winners, while historical-comparative studies (Nohlen 2009) analyse the development over time as an incremental reform process. In single case studies or small-N-designs, they reconstruct the initial conditions and actor constellations that led to a reform of specific content. They usually go into the details and developments of the cases and thus stand in contrast to the axiomatic parsimony of rational choice explanations. However, the ahistorical generalisations of the latter often only tautologically "explains" long-term change from altered utility maximisation calculations of the elites involved. Institutionalist explanations, which can be combined with both approaches, take a mediating position (Rahat 2011;Remmer 2008).
The vast majority of studies does not take into account the varying levels of democratic openness in a reform. 2 Among the exceptions is the policy cycle model of Norris (2011), which also mediates between theoretical generalisation and historical embedding. It helps structuring the focus on the long-term policy-making process in electoral reforms and avoids the unrealistic assumptions and tautologies of rational choice models without foregoing advantages such as the parsimony of assumptions with regard to political rationality (vested interests), for instance. It adds a realistic sequential order to the catalysts and barriers model of Rahat and Hazan (2011). Regarding the level of democracy as a context variable, the hypothesis can be derived from this perspective that democracy makes a difference for the processing of information from civil society and experts, thus, strengthening the bottom-up perspective of policy making. A higher degree of democracy and the resulting freer flow of information can effect on the entire cycle, but could be particularly evident in the problem definition, policy formulation, and the evaluation by experts and the public.
Norris' model reconstructs the policy cycle of electoral system reforms in four steps: a) agenda-setting, b) policy-making, c) implementation and d) evaluation and feedback. Policy analysts often add a stage "policy adoption" between b) and c), as we do in Table 1. The theoretical starting point is not an assumption about a reform actors' interest or goal, such as seat maximisation, but a concrete problem out of which a policy cycle emerges. In this way, the model approximates the historicalcomparative approach, but remains more structured. Similarities and differences already at the phase of problem identification as the starting point pave the path for possible developments, explaining incentives for institutional change also for former winners under a given electoral system. 3 The repeal of a reform as in the 2 Also Renwick (2010, p. 3) focuses largely on reforms in consolidated democracies. In contrast, unconsolidated democracies are in danger of institutional or systemic instabilities that might affect the electoral system, too. From the main types of Renwick (2010), the Turkish reform drive comes closer to the eliteimposed model, whereas for Israel, elite-mass interaction appears to be the more appropriate type. 3 Particularly in Rahat's (2006) study on the abolition of the Israeli reform, a proximity to Norris' policy perspective becomes visible. case of Israel requires for instance a well-informed public and freedom to change one's position. Table 1 summarises the adaptation of the policy cycle to the analysis of electoral reforms. This heuristic can serve as a framework of comparisons. While providing a sequential stage and multiple actor conceptualisation of electoral reform can be seen as an advantage of this model, it needs to be adapted to the peculiarities of this specific policy area. In contrast to comparing policy areas such as health care, housing, or defence (Rose 1973), in the field of electoral system reform, the selfinterests of the actors (the power dimension) is more likely to dominate the factual dimension. The policy perspective also provides analytical concepts such as policy networks and advocacy coalitions (Sabatier 1998) that can add up analytical capacity.

Directly elected chief executives in Israel and Turkey
Israeli-Turkish relations have recently attracted much scholarly attention (Efron 2021;Valansi 2018;Goren 2012;Quandt 2011; Sever and Almog 2019). However, comparisons of institutions (Heper and Itzkowitz-Shifrinson 2005;Sezgin 2001Sezgin , 2003 and policy making (Rivlin 2003;Yazgan and Yilmazkuday 2007) are much rarer, with the exception of the "politics of religion" (Tepe 2008 andSarfati 2009 andRubin and Sarfati 2016;Rubin 2020). Golan-Nadir (2022) justifies the case selection with the combination of similarities and differences. For instance, both "have long been considered as (the only) two Middle-Eastern democracies, although both have their share of shortcomings in substantive democratic elements, with strong orientation towards the West" (ibid., p. 9). Both countries also share a sense of "otherness" towards their Arab neighbours and have state founding ideologies (Zionism and Kemalism) that "relied on Judaism and Islam, respectively, to define a common denominator for their respective national identities." (ibid., p. 11) From the beginning, both have had to deal "with large ethnic/national minority groups-Arabs and Kurds, respectively" (ibid.: 10). The impact of the resulting cleavage reaches from party system fragmentation to security issues in both cases. Their special characteristics can be understood better and can contribute to theory development when analysed in comparative perspective (Hazan et al. 2021, p. 5).
The differences are primarily found in "political culture" (Golan-Nadir 2022). In Israel, social, religious, and political preferences are much more likely deliberated publicly in social media, interaction and via surveys. In Turkey, "public preferences are expressed mainly via personal behavior," (ibid., p. 11) Especially deliberation on religion-related issues is preferably kept private as they are perceived as sensitive issues. The differences continue when comparing the rise of religious-nationalist parties (Sarfati 2009), and the resulting challenge for the separation of powers. Hazan (1996) has characterized the reformed Israeli system as "presidential parliamentarism", whereas in Turkey, a "loaded discussion on system change" (Akgün 2016) is still ongoing, as opposition parties joined their efforts for an enhanced parliamentary system (Adar and Seufert 2021).
In both cases, voters could concurrently vote for the head of state, respectively government (a prime minister in Israel and a president in Turkey), and for a list of candidates to be send into parliament. For the former, a Two Round System was chosen in both cases. However, the Israeli prime minister was required to ask for a vote of confidence for his cabinet in the Knesset, while the directly elected Turkish president, since he became also head of government in 2018, does not face any restrictions for choosing his cabinet ministers and vice president(s) (Adar and Seufert 2021). In both cases, the reforms were meant to facilitate government formation. In practice however, the outcome in the Israeli case did not meet the expectations and the reform was soon reversed. In the Turkish case, with the constitutional reform of 2017 the direct election became centrepiece of further autocratisation.
Although the Israeli reform had initially cross-party support, it coincided with the rise of Likud in the early 1990s. Similarly, the reform in Turkey must be seen in the context of the consolidation and expansion of power by the AKP under Erdogan. There are parallels even in the power structure of the long-time ruling coalition formations, with a dominant right-wing party (Likud resp. AKP) supported by nationalist or religious parties such as the smaller ultraorthodox parties in the one case and the nationalist movement party (MHP) and the Great Unity Party (BBP) in the other. However, at the time of reform, the fragmentation of the Israeli party system was much higher than that of the Turkish one. Both have had prior K experiences with such a system at local level; in Turkey mayors are directly elected by plurality, in Israel a similar system is used since 1978. 4 Methodically, this study applies an asynchronous comparison that could make it more difficult to control for external influences. However, it is also possible that similar developments such as female suffrage or nationalist movements take place at different points in time in different countries. Mylonas and Shelef (2017) stress that the timing when similar events take place might be influenced by similar causal processes despite varying points in time. Asynchronous comparisons are vindicated when "the research question and the research design require certain initial conditions to pertain before a causal effect can materialize. On top of this, showing that an argument operates in a similar fashion in two distinct periods in different contexts comes with many of the virtues of a most different systems design" (ibid., p. 152).
Some fields of study such as (stateless) nationalist movements or woman's movement heavily rely on asynchronous comparisons. In this study, both cases show regional and temporal proximity, and both reforms initially had similar justifications in the public discourse. It is the different trajectory of each reform that enables to explore differences or "causal effects" of political contexts. Mylonas and Shelef (2017) argue that an asynchronous comparison enables a leverage of differences between the selected cases.

Research question
Research of electoral reform usually focuses on democratic systems and, more rarely on forms and functions of elections in authoritarian cases (Schedler 2015). The bulk of literature does not take into account different levels of democratisation as relevant for the outcome of an electoral reform. The focus on "elections as instruments of democracy" (Powell 2000) even risks to neglect their (limited) role in authoritarian systems. In Gallagher and Mitchell's (2008) comprehensive case selection, only the Russian case study systematically addressed this topic. Norris (2004, p. 15) briefly points at the role of elections in post-authoritarian transformation, and also Rahat and Hazan's (2011) barrier model does not directly address this as a context variable.
As there is no accepted theory of how varying levels of democracy might influence electoral reform, advice can be drawn from similar studies of policy change. Analysing budgetary decisions, Baumgartner et al. (2017) assume that authoritarian regimes may operate with fewer institutional barriers and have less incentives to collect and respond to relevant civil society information. They might respond more quickly to shifting contexts than democratic ones, which provides them with "institutional efficiency" to react. 'Major' electoral reforms (Jacobs and Leyenaar 2011) could thus be facilitated. On the other side, democracies with their checks and balances and their need for majorities to change the status quo might respond more slowly, but can have a better capacity to collect and process information about social and political preferences "because of mechanisms associated with electoral accountability, as well as stronger and more independent civil society organizations including the press", giving them an "informational advantage [...] about the impact of current policies through a more vibrant network of civil society organizations, including political parties staffed by officials anxious to 'feel the pulse' of various constituencies." (ibid., p. 793) Regarding our cases, this leads to the assumption of a comparative advantage of the Turkish system in "institutional efficiency" to react, while Israel's better democratic performance points at a comparative advantage in information processing and inclusive responses.
The informational advantage hypothesis of higher democratic scores is in line with the principal-agent theory (Gailmard 2014), which points for instance at the problem of informational asymmetry between principal and agent and how the former can develop effective institutional control practices of the better-informed agents. In other words, better democratic control mechanisms make the exploitation of information asymmetries in electoral reform by the government more difficult. Against this backdrop, we expect that a better democratic performance can make rapid reforms more difficult, whereas higher levels of autocracy facilitate electoral reform in case the regime is unpleased with the status quo. Better democratic checks and balances might hinder populist "strongmen" to paint their own electoral or even constitutional system. Rising autocracy should find expression in a reduced number of institutional and partisan veto players (Tsebelis 2002). In the Eastern Mediterranean region, Israel and Turkey provide varying political conditions for a similar outcome, the introduction of direct election for a chief executive, albeit at different periods in time. This article explores factors that help explaining first the introduction of this reform and later their different trajectories.

The democratic environment of both cases
In order to prepare focusing on the democratic environments, this section depicts data from the V-Dem data set. The Varieties of Democracy (V-Dem) project aims to grasp different conceptualisations of democracy, such as an electoral, liberal, deliberative, egalitarian, and participative variant. It is one of the most advanced instruments in this field due to its elaborate and granular structure (Coppedge et al. 2021a) and has become very common in studies of democratic deconsolidation and regression (Schäfer and Zürn 2021). For our purpose, the liberal democracy index (LDI) is particularly interesting. It comprises two components: the electoral democracy index comprises indicators of free and fair elections, and the liberal component index such indicators as rule of law, respect for civil rights, and restraint of the executive by the judiciary and legislature. In addition to the LDI, Table 2 depicts figures for performance legitimation, high-court independence, rational-legal legitimation, and civil-society participation of Israel and Turkey.
The LDI ranges from 0 to 1000, with 500 to be considered as a threshold between democracy and autocracy (Schaefer and Zuern, p. 172). In Turkey, liberal democracy peaked in 2004 and declined since, while Israel's scores are stable in the long term. The biggest drop for Turkey was in 2014 to 263 (-87). This downward trend, also visible in other variables, coincides with events such as the Gezipark protests (2013), the beginning presidentialisation (2014), the 2015 repeat election (Sayarı 2016), and the government crackdown after the July 2016 failed coup attempt. In 2020, while  IL  TR  IL  TR  IL  TR  IL  TR  IL  TR   1990  685  368  2167  2636  1289  1555  1372  579  883  456   1991  684  379  2167  2636  1289  1555  1372  579  883  456  1992  680  391  2167  2636  1289  1555  1372  579  885  476   1993  680  391  2167  2636  1289  1555  1372  579  885  476   1994  688  392  2167  2636  1289  1555  1372  579  885  476  1995  688  392  2167  2636  1289  1555  1372  579  919  Israel was already in its snap election cycle and Turkey began discussing its next electoral reform, the gap increased to 537 points. These figures quantitatively illustrate the varying democratic contexts of the electoral reforms under consideration and help explain the different dynamics of the electoral reform. At the time of the introduction and abolition of the Israeli reform (1992 and 2001), its LDI score was about 680, while the Turkish score in 2012 was 384, a difference of nearly 300 points.
The indicator performance legitimation 5 serves as a proxy to evaluate the institutional-efficiency conjecture, as introduced above. It should be noted that this indicator, like rational-legal legitimation below, aims to capture these claims from the government's (not the population's) point of view by means of expert surveys. The two indicators thus show how strongly a government relies on these two aspects in its quest for legitimacy. However, the indicator shows whether or to what extent a government sees this dimension as one of its strengths. In this variable, Turkey scored well until 2007, when the values began decreasing, but still above those of Israel until 2015. Turkey's strength in this variable supports our initial hypothesis of a comparative advantage of less democratic systems in aspects of institutional efficiency-at least from a government's point of view.
Until 2013, V-Dem estimated high court independence in Turkey better than in Israel (Table 2,  . For our purposes, it is crucial that this indicator gives higher values for Turkey than for Israel up to 2013. The indicator suggests that Turkish courts were at least as "free" as Israeli courts from political influence to make decisions. For instance, in the 2007 presidential election crisis, the opposition first boycotted the vote in parliament and then questioned its validity by calling the constitutional court, which then annulled the vote. President Erdogan's reaction was to call for early parliamentarian elections and to submit a reform package including the direct election of president. This time the constitutional court had no objection, and the reforms were accepted in a popular referendum in October 2007. Despite the democratic decline in Turkey, its rational-legal legitimation 6 surpassed the Israeli scores from 2003 to 2009. This points at a temporarily buildup of institutional efficiency to react as expected above. However, it must again be taken into account that the estimations of the government side were measured here. The subsequent steep slope suggests that this indicator was not separated lastingly from the overall democratic regression. Discontent with the politicisation of the presidency is shared among the parties that strive for an augmented parliamentary system. The mismanagement of the current economic crisis further contributed to the decline of the changed system's legitimacy. In contrast, perceived legitimation claims in the Israeli case were stable over the whole period. This first quantita-5 The V-Dem Codebook (Coppedge 2021b, p. 222) defines it as the extent in which the government refers "to performance (such as providing economic growth, poverty reduction, effective and non-corrupt governance, and/or providing security) in order to justify the regime in place". 6 The V-Dem Codebook (Coppedge 2021b, p. 222) defines this variable as the extent to which the current government refers "to the legal norms and regulations in order to justify the regime in place?" It "pertains to legal norms and regulations as laid out for instance in the constitution regarding access to power (e.g. elections) as well as exercise of power (e.g. rule of law). Electoral regimes may score high on this question as well as non-electoral regimes that emphasize their rule-boundedness." (223). tive inspection supports our initial conjecture that Israel is doing better in areas of democratic checks and balances, while Turkey has had (temporarily) advantages in institutional efficiency. 7 Finally, civil society participation was chosen from V-Dem to test the conjecture of a better information flow in more democratic countries. This correlation was derived from the informational-advantage hypothesis. While the values for Israel show some decline since 2001, it remained always well above the Turkish scores, where the decline was more gradually from the peak in 2015. The biggest drop was then in 2016 with -114 points, which corresponds with the purges after the July 2016 failed coup attempt.

The policy cycle of electoral reform
The policy cycle model as adapted in Table 1 is now applied to facilitate a structured, focused qualitative comparison (George and Bennett 2005) of the reform in both cases.

Agenda setting
Agenda setting starts with a problem definition: What is described by whom as a problem (for whom) and why? In both cases, coalition bargaining and formation was perceived as intransparent, dominated by a spoils mentality among the parties and often short-lived coalitions (Kalaycioglu 2016). Even though the problem identification sounded similar in both cases (governance problems and cabinet stability), there are clear differences when the defining actors are taken into account. In Israel, attempts for electoral reform were much older, but in the early 1990s they gained significant support and reached a consensus on the adoption of a directly elected prime minister with the Basic Law: The Government (No. 1396) in spring 1992 under the premiership of Yitzhak Shamir (Likud). However, Likud did not support the reform, while individual members such as Netanyahu did. The cross-party reform initiative aimed to address rising fragmentation and polarization among the Knesset parties, contributing to government instability. More precisely, the reform was motivated by a government crisis in early 1990 that saw "an unprecedented public orgy of floor-crossing and unseemly bargaining, with parties and individual legislators scrambling for place, preferment, and political advantage." (Ottolenghi 2001, p. 109).
Turkey's electoral system has been frequently subject to changes (Hale 2008). Probably its most controversial feature is the ten percent nationwide threshold, which was introduced in 1982 and first applied in the November 1983 parliamentary 7 Both Israel and Turkey have a unicameral parliament, leaving the state president and the supreme resp. constitutional court as institutional veto players. In Turkey, the state president was neutralised as veto player in 2007 with the election of Abdullah Gül from the governing party; later the office became central for the "presidentialisation" of the Turkish system of government. The Israeli state president is a largely ceremonial figure, elected with qualified majority by the Knesset. During the 1992 reform, Chaim Herzog (Labour) was president. Labour supported the reform. election. Already in 1961, the system was changed from plurality in multi-member districts to a PR system (d'Hondt) with a threshold at district level (Table 3). The extraordinary ten percent threshold was meant to enhance government stability after the experiences of violence and political instability in the 1970s. Its effect of reducing the number of parties in parliament was particularly strong in the 2002 elections, when only the newly founded AKP and the CHP managed to pass. Its "psychological effect" influences voters' choices in favour of those parties that are expected to pass the hurdle, or to form alliances or to run their candidates as independents, in which case the threshold does not apply (Hale 2008). Cop (2011, p. 5) went even further and observed an "extremely volatile character of electoral system changes that took place in Turkey from the end of the Second World War up until the end of the 1990s." He identifies the desire to prevent the repetition of past experiences and short-term goals of politicians as reform motives.
The problem definition and agenda setting took place during the 2007 presidential election crisis, when the election of the AKP candidate was boycotted by the opposition in parliament and the general staff tried to influence the course of events by a statement in the election night. In response, the government called new parliamentary elections in July. A clear victory of the ruling party enabled their candidate to be elected in August 2007 in the third round, with only an absolute majority required. As a consequence of this crisis, the AKP put the direct election of the president on the agenda.
The difference in this first phase come out clear. In Israel, agenda setting happened bottom up by a broad coalition of civil society and party-political interests, whereas in Turkey it was done top down by the ruling elite. While in Israel, it was more an incremental process, in Turkey it was very much event-induced. However, it both cases, the problem definition encompassed arguments that the "will of the people" has been distorted by a) procedural requirements of a presidential election by parliament as in the Turkish case, or b) over-proportional influence of small parties on government formation, stability, and policies as in the Israeli case.

Policy making
The differences that became visible in the first phase also influenced the second phase. Accordingly, in Israel policy making had to pay much more attention to creating a majority in parliament than in Turkey. Thus, more responsiveness had to be developed toward smaller interest groups and parties. Factors such as democratic openness, the availability of information contribute to plurality of opinions and demands at this stage in this process. In Israel, there were extensive revisions of the initial proposal at this stage, resulting from the need to establish a parliamentary majority.
The initial plans were watered down in the parliamentary process. For instance, the principle that Knesset and prime minister elections must always take place at the same day was evaded by the special election of the prime minister in case of his/her resignation. The initial idea of linking the tenure of Knesset and prime minister was also labelled as "the balance of terror": "[I]n case of a no-confidence vote, parliament must also always go to early elections. In case of early dissolution initiated by the prime minister, the prime minister always resigns and faces again the popular test." (Ottolenghi 2002, p. 96 f.) The challenge to find sufficient support for the 1992 reform let the initiators try to meet diffuse and heterogeneous expectations, in order to mobilise sufficiently broad support. The reform attracted backing especially by the prospect of taking away the choice of chief executive from the small and extreme parties and granting it to the voters, which "would confer upon the prime minister both a mandate and legitimacy." (Rahat and Hazan 2008, p. 345).
In contrast, electoral reform policy making in Turkey was centrally controlled, without much need and time for compromise or even civil society consultation. After the election of a new president by parliament in May 2007 failed due to opposition boycott, the AKP immediately drafted a bill to reform the election procedure, including the direct election of the president, enabling a second term, and reducing the parliamentary quorum that is need for parliamentary decisions. The bill was adopted in May 2007, but president Sezer vetoed it, and it was adopted a second time by parliament, thus overriding the veto. Sezer then referred it to a popular referendum in October 2007, too late to take effect in the 2007 presidential election. The policy making was thus divided into two steps, a quickly formulated first step for the 2007 bill and referendum, and the detailed elaboration that followed in 2012.

Policy adoption
Policy adoption is influenced by various internal and external factors, such as the parliamentarian's and party's expectations of costs and benefits. It is striking that in both cases no direct costs were to be expected with regard to the parliamentarians' own re-election, since the regulations to vote for parliament was not affected. However, this factor played a greater role in Israel than in Turkey, where due to the AKP's dominance, majorities in parliament could be taken for granted. Related to this are different positions of the two cases on the axis of executive versus legislative dominance. Israeli governments usually comprise a high number of parties with varying degrees of party cohesion and ideological congruence between the parties. In the 1992 reform, this was however less relevant, as it was supported by a crossparty initiative. Neither the State President nor the Supreme Court made use of their veto power. 8 The Turkish case can be summarised as the art of neutralizing veto players. With the succession of Abdullah Gül to the Kemalist Ahmet Necdet Sezer in 2007, executive dominance was further strengthened. As in the Israeli case, it was a separate regulation and did not change the procedure of electing the parliament. The reform proposal was accepted by 68.95% in October 2007 in a constitutional referendum. The reform details were adopted in the 2012 Law on Presidential Elections (Cumhurbaskani Seçimi Kanunu) by the Turkish parliament on 19th January 2012 (Act No. 6271, Resmi Gazete of 26th January 2012, No. 28185). It adopted new rules and procedure how the president is to be elected. The presidential term was reduced to five years again and a second term in office was enabled. Even a third term was made possible, if parliament called for snap elections during the second term of a president. (Umit 2015).
To sum up this stage, the Israeli process of reform adoption was legislatively dominated, while the reform process in Turkey was dominated by the executive. It was a process driven by the government with no civil society participation or even alternative, competing proposals. While the use of popular referendums is not an uncommon strategy for the ruling party, in this case it has been the then president who called the referendum.

Implementation
This phase offers fewer surprises. In Israel, the 1992 amendment to the Basic Law: The Government was implemented with the next elections in May 1996, with Binyamin Netanyahu being elected prime minister in the first round with 50.4% of valid votes against Shimon Peres with 49.5% (Hazan 1997). However, it also turned out that the expectations of the reform agents for "straight ticketing" did not materialise. Instead, voters discovered and used the opportunity to split their votes between different parties in the Knesset and the prime ministerial election. Especially smaller parties benefited from this development, and the fragmentation of the Knesset rose to an unexpected peak.
In Turkey, the first direct election was implemented in August 2014 (Erdogan 2015), and saw former prime minister Recep Tayyip Erdogan winning with 51.79% in the first round against a cross party opposition candidate and a HDP (pro-Kurdish) candidate. Erdogan's direct election as president was a significant step in the majoritarian transformation of Turkey's system of government. Shortly thereafter, for example, the president's neutrality requirement was lifted to allow Erdogan's party membership to resume. Another consequential problem was the "dual leadership" of president and prime minister, which was eliminated in a constitutional reform in 2017. With it, the position of prime minister was dissolved in 2018 and its competencies transferred to the directly elected president as part of the "presidentialisation" of the Turkish system of government (Akman and Akçalı 2017;Krumm 2018;Özsoy Boyunsuz 2016).
At implementation stage, the constitutional court later clarified that also previous presidents have a right to stand for a second term (Umit 2015, p. 174). Each party was allowed to endorse only one candidate with at least 20 MP's signatures, disabling smaller parties to run their own candidate (ibid.). In the presidential election of August 2014, registered Turkish citizens abroad could cast their votes in embassies and consulates around the world for the first time. Especially in Western European countries, the AKP achieved exceptional results.

Evaluation and feedback loop
At this stage, significant differences become visible again. In both cases, "ticket splitting" led to unintended consequences. In Israel, parliamentary fragmentation rapidly increased, requiring even larger coalition formats, while in Turkey the reform created different dynamics for parliamentary and presidential elections. It detached support for Erdogan from support for the AKP. Thus, in both cases, the prime minister had much more problems putting together a working majority in parliament. In parliamentary systems, effective governance "cannot be achieved without a loyal parliamentary majority." (Ottolenghi 2002, p. 97) In presidentialism, this problem is well known as "divided government". In Turkey after the June 2015 elections, this split even ran through the ruling party. In Israel, the reform alone was not able to transform the structure of competition to two alternative candidates (or party blocs). The problem is well known also in the Turkish case, where a subsequent reform of the parliamentary electoral law in 2018 introduced the option of pre-election alliances as an opening mechanism to the unpopular 10% threshold.
Contrarily to the introduction, the Knesset's 2001 reverse reform was almost uncontroversial and smooth. Some minor changes were adapted; most notable, a "semiconstructive" vote of no confidence (Lento and Hazan 2021) that could replace a prime minister ( § 28), turned into a "full constructive" one (Friedberg and Hazan 2021, p. 308). This was also meant as a compensation for the sitting prime minister, as he had (again) to put together a parliamentary coalition prior to his election (Mahler 2016, p. 162). Until 1996, a simple majority of those present and voting in the Knesset could oust the sitting prime minister by a vote of no confidence.
Both the 1992 reform as well as the 2001 adoption of the constructive vote of no confidence inserted a majoritarian and personal element into the political system, in addition to the primaries at the level of party organisation by several parties (Rahat and Hazan 2008, p. 342). 9 The 1992 reform was introduced under the right-wing Shamir government , that initially comprised twelve parties, its abolition took place under the first Sharon cabinet in 2001, comprising seven parties that formed a surplus coalition. Despite himself being directly elected with large majority, Sharon endorsed the reversal of the reform immediately.
In 2020, the idea of a direct election of the prime minister re-emerged in the course of the four inconclusive elections between April 2019 and March 2021. As an attempt to overcome the impasse, in April 2021, a bill proposing direct election of the prime minister was submitted by two MKs from the ultraorthodox Shas party. This was widely seen as an attempt to rescue the premiership of Netanyahu. The bill proposed a single election only for the office of prime minister, who would then head a caretaker government with the task to form a governing coalition in parliament. If he fails to do so, another Knesset election would be triggered. For this bill, an absolute majority in the Knesset would have been needed. Polls showed Netanyahu despite his indictment of corruption charges as the most popular politician at this time. 10 Israel is still used to fragile multi-party coalitions. In 2021, the "rainbow coalition" comprised eight out of the 13 parties in parliament. As shown in Table 4, party system fragmentation at electoral and parliamentary level jumped to unprecedented highs in the 1990s. With the reverse reform in 2001, fragmentation did significantly fall (2003), but remained still well above the pre-reform level.
In Turkey, a recent reform was the opening up of the threshold for party alliances prior to the 2018 parliamentary election. Evci and Kaminski (2021) argue that this was an "ex-post mistake" by the governing party (ibid., p. 482). 11 According to their simulation, the AKP would have done much better in the 2018 parliamentary elections regarding their seat share under the old system, because the nationalist 9 The system was then stable until the government crisis commencing 2019 with its four consecutive elections in two years. As a solution to this crisis, the title of designated or alternate prime minister was introduced by the Likud-Blue and White coalition in 2019, and also applied by the Bennett-Lapid government (eight parties) in 2021. 10 Thus, it is not surprising that the idea of direct election was circulated already after the second inconclusive election in September 2019. "You say the public doesn't want me, is tired, is vomiting me out. Let the public decide!" Netanyahu said in December 2019. See Jerusalem Post, Dec. 6, 2019, p. 1: Netanyahu: Direct election for PM "interesting". 11 As an ex-post mistake, they (vaguely) define electoral engineering that turned out to be backfiring for the initiating party(ies). The idea of ex-post mistakes in institutional engineering comes out clear also in the observation of Adar and Seufert (2021, p. 35) that the new governance system "has produced anything but encouraging results for the AKP". For instance, they point at the growing influence of the junior alliance partner, the nationalist MHP, in a range of policies and in public service recruitment. In the pre-AKP era, electoral reforms were often initiated after military coups to address the (allegedly) shortcomings and malfunctioning of the pre-coup polity. However, the use of a PR system within parliamentary government was never in question. While turnout at the ballots in Turkey is high, the post-election bargaining to form coalition governments that emerged since the 1970s was unpopular. This experience was used as one of the rationales of the 2018 reform, which set incentives to form electoral alliances between parties before an election, not afterwards.
While the unpopularity of the partisan "horse trade" after elections was a characteristic in the pre-AKP era, the problem was factually (not institutionally) eliminated by the AKP supermajorities in parliament since 2001. It reappeared in the June 2015 election crisis, when the AKP under then prime minister Ahmed Davutoglu lost its absolute majority in parliament for the first time (Çarkoglu and Yıldırım 2015).
Another reform element was the enabling of pre-election alliances in 2018. As a consequence, two alliances exist today: The People's Alliance comprises the AKP, MHP, and BBP (Great Unity Party), while the Nation's Alliance comprises the republican CHP, Good Party, Democrat Party (DP) and Felicity Party. The Kurdish based HDP has not joined, but cooperates. As breakaways from the AKP, Ahmet Davutoglu's Future Party and Ali Babacan's Democracy and Progress Party (DEVA) are expected to join the Nation Alliance sooner or later.
In early 2021, the AKP and MHP held separate deliberations about the next reform step of the Election Law and the Law on Political Parties, which was regarded as in need of an adjustment to the presidential system of government. Main points are the lowering of the electoral threshold, preventing lawmakers to switch their parties, and provisions regarding the participation in elections. The 2021 reform debate also considered smaller multi-member districts or even a shift to a singlemember plurality system. This would require the division of the current 86 districts into 600 single member districts.
To sum up, for the introduction and abolition of the reform in Israel, the public discourse and available information played a different role than in Turkey for the follow-up reform of 2018. The higher level of democracy in Israel is then also reflected in the lower coherence of the camps, which made it easier to change sides or opinions (Rahat 2006, p. 44). In Turkey, on the other hand, the whole process of reform is much more centralised and controlled by the government, the flow of information is regulated and the camps stand antagonistic to each other, making it much more difficult for actors to change sides-or minds. Under the Erdogan governments in Turkey, power is increasingly centralised, making it much easier to quickly push through reforms once the formation of will in the ruling party is completed. Through the "pooling" of media and civil society, the flow of information and opinion can be largely controlled, although this does not directly translate into legitimacy. However, even in this more power-centralised case, the timing of a reform can best be explained by situations of political crisis for the ruling elites. Here, the presidential election of 2007 and the parliamentary election of 2015 could be identified as catalysing events that put the issue of electoral reform (respectively direct election of the head of government) on the agenda for the AKP.

Conclusion
This study compared the process of introducing, legitimating and (not) maintaining the direct election of the chief executive in Israel and Turkey within the framework of a policy cycle model for electoral reform. The model offered a valuable framework for comparison; only in the last phase it became difficult whether some reform proposal should be seen as part of the feedback-loop or as already agenda-setting of a new policy cycle. The two cases illustrate how different levels of democratisation can contribute to different trajectories of electoral reform. Depending on the political context and the number and strategies of the actors, a similar policy can take on different trajectories. Table 5 summarises some of the results.
Derived from the informational-advantage hypothesis, institutional efficiency effects appear more likely in more autocratic cases, and accountability advantages in the more democratic ones. The ongoing majoritarian reform process in Turkey points exactly in this direction of aiming to enhance institutional efficiency (from a top-down perspective), at the costs of accountability and information processing by (parts of) civil society. The case of Turkey shows, however, that this advantage was not permanent; democratic decline was soon followed by institutional inefficiency (Table 1).
In the Israeli case, policy makers were listening to the communicated problems and reversed the reform in 2001 (Rahat 2006(Rahat , 2008. The results of the case studies suggest that information processing is more inclusive in more democratic cases and more government driven in the less democratic ones. This makes it easier to respond to the unexpected outcome of increased party system fragmentation (due to strategic voting) in Israel, while in Turkey the reform was subject to intra-elite considerations and became part of a broader presidentialisation strategy of these very elites.
In the Israeli case, the reform was a majoritarian element in a predominantly proportional-minded framework of politics, while in the Turkish case the directly elected president is a majoritarian element in a traditionally majoritarian framework K  (Lord 2012). Furthermore, political culture in Turkey is much more prone to rapid institutional change, while institutional change in Israel occurs much more gradually. At technical level, a lesson from the Israeli experience of the 1990s was that direct election can sustain only if it comes with an electoral system that also enables strong parliamentary majorities for the directly elected premier (Ottolenghi 2001, p. 121f.) While the Turkish "alliance model" points in this direction, it also illustrates the strengthening of the executive ("rule by decree", see Yilmaz 2020) at the costs of parliamentary control. Ottolenghi even anticipated "the risk of a populist or even a plebiscitarian-authoritarian turn" (ibid., p. 121).
With the reverse reform in Israel, the PR system dating back to the pre-state period remained dominant. In Turkey, the reform was an intermediate step in the process of presidentialisation. This also created a need for readjustments to the interests of the ruling party, such as to address the "problem" of breakaway parties and splinter groups, of kingmakers and spoilers. Government stability was enforced at the price of checks and balances, while in the Israeli case the system of checks and balances led to a more cautious "evolution" of the system.