Does the Electoral System Foster a Predominant Party System? Evidence from Turkey

This paper discusses the role of the electoral system in making the Justice and Development Party (AKP) dominant. Drawing on Sartori’s framework, we first clarify the concept of a predominant party system. Second, we examine the impact of the electoral system on the emergence of a predominant party system in Turkey. Analysing election results, we argue that the electoral system fosters dominance in three ways. First, a combination of electoral formula, national threshold and district threshold leads to over-representation of large parties and underrepresentation of small ones. Second, the fear of a wasted vote due to the high threshold prompts voters to support their second-best option, which concentrates the votes among large parties. Finally, the electoral system increases electoral turnout rates by extending polarization.

ones. This frequently promotes one-party governments. Second, the electoral system leads to the concentration of votes among large parties due to the fear of a wasted vote. In this context, small parties have no option but to ally with other parties or run with independent candidates. Third, the electoral system fosters electoral turnout by stimulating polarization.
In Section 1, we review different operational definitions of predominant party systems and explain why this is an appropriate categorization of the Turkish party system in its current form. In Section 2, we briefly evaluate different theories of the emergence of the predominant party system in Turkey. In Section 3, we use institutionalist theory to explore the role of the electoral system in shaping the Turkish party system. The last section concludes.
Orthodox definitions of a dominant party take into account two variables: time in power and size of vote/seat share. Virtually all scholars agree that dominance is established over time; 4 however, the threshold they specify varies considerably: a single reelection (van de Walle and Butler 1999); three elections (Sartori [1976(Sartori [ ] 2005; 20 years, four consecutive elections or one generation (Blondel 1968;Greene 2007); or 30-50 years (Cox 1997). 5 Regarding the vote/seat share, on the other hand, the criterion ranges from an absolute majority in parliament (Sartori [1976(Sartori [ ] 2005Ware, 1996) to a plurality of votes and seats (Duverger 1959;Pempel 1990).
Drawing on Sartori, we suggest that a dominant party is one that wins at least three consecutive elections in a competitive political environment with more than a 10 per cent margin and plays a major role in government formation. We add two criteria to Sartori's framework, justifying its inclusion as follows. First, we contend that a doubledigit margin over the second party is sufficient to qualify as dominant party. 6 Second, "the major role" in government formation indicates that the dominant party forms the government alone or with the support of minor parties, i.e., when it forms minority governments. 7 This definition covers the Swedish Social Democratic Party and the Christian Democrats in Italy, which were the biggest partners in coalition governments or formed minority governments, as well as the AKP (Turkey) and the Indian National Congress (INC), which formed government alone.
A dominant party is a necessary but not a sufficient condition for the emergence of a predominant party system, however (Sartori [1976(Sartori [ ] 2005. To transform the party system into a predominant one, the dominant party must form the government alone. This criterion is crucial, since it disqualifies, for instance, Sweden (moderate pluralism) and Italy (polarized pluralism) where the dominant parties could not form government alone. Meeting these criteria, the universe of cases with predominant party systems includes, but is not limited to, Mexico, Taiwan, India, Japan and Turkey.
For the sake of conceptual clarity, predominant party systems must be distinguished from the concept of the hegemonic party system. First, unlike the former, hegemonic party systems are essentially non-competitive. In this sense, the hegemonic party will remain in power whether "it is liked or not" (Sartori [1976(Sartori [ ] 2005. 'Other parties are permitted to exist, but as second class, licensed parties as they are not permitted to compete' (Sartori [1976(Sartori [ ] 2005. Second, building on the first aspect, a hegemonic party is more strongly represented in parliament than a dominant partyit would be able, for example, to unilaterally change the constitution (Magaloni 2006: 35). Third, electoral malpractice is pervasive in the hegemonic party systems. This may be ascribed to the unilateral control of organizing, monitoring and adjudicating elections by the hegemonic party (Magaloni 2006: 36).
Since 2015, the Turkish party system has shown some signs of moving towards a hegemonic party system (C ß ınar 2016; Lancaster 2016; € Ozbudun 2015). First, events between the June and November 2015 elections showed that the incumbent party would resort to any means to keep its power. Second, evidenced by allegations of fraud and intimidation in the April 2017 referendum on moving to an executive presidential system, suspicion of electoral malpractice has become more frequently aroused. Moreover, the use of repressive powers against the opposition has become more pervasive since the Gezi Park protests. The declaration of a State of Emergency after the failed coup attempt in July 2016 further legalized repression of the opposition. The third element, the position of the opposition parties vis-a-vis the AKP, is more debatable. Although the ultra-nationalist Nationalist Action Party (MHP) has apparently being acting as a satellite of the AKP since July 2016, possibly due to intra-party contestation, it would not be fair to put the rest of the oppositional parties in the same basket.
These changes, the blurring of the line between the party and the state, have significant implications for the nature of the regime. 8 It is widely documented (e.g., Esen and G€ um€ us ßc ß€ u 2016; € Ozbudun 2015) that the regime is moving towards "competitive authoritarianism", in which the incumbent party skews the political field to its advantage 7 The analytical distinction between the dominant party and the predominant party system is adopted from Sartori. In his classification, the Italian Christian Democrats, the Israeli Mapai, or the Danish Social Democrats were dominant parties as they outdistanced their rivals and remained in power for decades (Sartori [1976(Sartori [ ] 2005. Unlike predominant party systems, however, in these cases, the dominant parties had to form (mostly shortlived) coalition governments with other parties. 8 This trend has been observed by international scrutineers. For instance, according to Freedom House (2018), Turkey moved from the category of a "partly free" country with the freedom score of 3.5 in 2015 to the category of "not free" with the freedom score of 5.5 in 2018.

How Do Predominant Party Systems Emerge? A Theoretical Perspective
Several theories have been developed to account for the emergence of predominant party systems. Table 1 briefly applies these theories to the cases we have identified. Each theory comes with drawbacks. For example, performance legitimacy theory, 10 which posits that predominant party systems emerge because of dominant government parties' remarkable economic performance, fails to account for the rising vote share of the AKP in the November 2015 elections despite deteriorating economic indicators since 2008 (Acemo glu and Uc ßer 2015; C€ omert and C ß olak 2014). 11 Similarly, the decision-centric model, 12 which assumes that dominant parties conquer the centre and thus represent the median voter (Arian and Barnes 1974;Cox 1997;Greene 2007;Riker 1976), does not fit the Turkish case, since the AKP is a right-wing party that incorporates neoliberalism, conservatism and selective nationalism in its political appeal. 13 Not all theories are irrelevant, however. For instance, once the AKP began to fear for its future, in 2008, it started to follow a polarization strategy, mainly along religious lines (C ß ınar 2016). This makes social cleavage theory relevant to the AKP case. 14 Similarly, the clientelist use of public resources matters with regard to the AKP's dominance (Akda g 2014; Aytac ß 2014; C ß arko glu and Aytac ß 2015; Kemahlıo glu 2012; Sayarı 2014). Neoliberal populism (Bozkurt 2013;Dorlach 2016) under a charismatic party leader is another element that fosters the AKP's dominance. 15 9 We must note that opposition victories in large cities in the March 2019 local elections constituted a major setback to Erdo gan's tightening authoritarian rule. Recently, Erdo gan's regime has also faced intra-party splits. Former Prime Minister Ahmet Davuto glu and former Minister of Economy Ali Babacan publicly expressed their intention to set up their own parties. These developments may be the beginning of the end for Erdo gan and his party, since small shifts in votes may cost them power. This conforms to Duverger's (1959: 312) arguments: "the dominant party wears itself out in office, it loses its vigor, its arteries harden. It would thus be possible to show . . . that every domination bears within itself the seeds of its own destruction." 10 While it can plausibly be argued that the Koumintang Party (KMT) in Taiwan and the Liberal Democratic Party (LDP) in Japan took advantage of prolonged economic growth, others, including the Institutional Revolutionary Party (PRI), were able to maintain their rule despite a deteriorating economic record (Greene 2007: 19;Magaloni 2006: 13). 11 Bas ßlevent et al. (2005) found that the AKP gathered huge support from those whose economic conditions deteriorated after the 2001 financial crisis. 12 Based on an expert survey, Huber and Inglehart (1995) positioned the LDP, PRI and the INC on the right with medium scores of 8.43, 6.20 and 5.80 respectively, on a scale that goes from 1=left to 10=right. 13 This is the case notwithstanding that the AKP arguably represents the median voter (Tuncer and Sa gdic ß 2017). 14 Despite its popularity, social cleavage theory has been contested on several grounds. Above all, social cleavage theory does not apply to all cases of predominant party systems, particularly India, where the INC achieved more or less equivalent support from different classes, ethnicities and religions (Chhibber and Petrocik 1989). Second, social cleavage theory may invert the direction of causality. Sartori (1990) contends that, even if it is true that societal changes are registered through political parties, it is also true that parties might (de)politicize issues that are key to the functioning of the system. Third, social cleavages are not static and might strengthen or weaken over time, especially when parties cannot find any new cleavage to mobilize voters (Ware 1996: 127); this has been the case in Turkey in recent years. Last but not least, parties not only reflect social cleavages, but also actively shape them (de Leon et al. 2009). 15 For instance, according to a Konda (2018) report on the profile of their voters, 46 per cent of AKP supporters primarily vote for the leader.
Electoral systems have commonly been viewed as durable institutions (Lijphart 1994;Taagepera 2007), due to requiring majorities or even qualified majorities in the legislative arena to make any change to them. On the other hand, parties in power try to alter electoral systems if they no longer serve their interests (Sartori [1976] 2005) or if they are unstable and fragmented (e.g., Italy, Israel) or hyper-stable (e.g., Singapore, Japan) (Norris 2004: 81).
Despite the significance of this issue, the relationship between predominant party systems and electoral system has remained underexplored (Erdmann and Basedau 2007: 10). 17 Indeed, the design of the electoral system is of greater importance in predominant party systems than other party systems, given that small changes in the electoral formula may translate into a big return for the incumbent party (Sartori 1990: 347).
In addition to exerting direct influence, the electoral system may also indirectly impact the party system. In this regard, it is widely suggested that electoral systems that favour personalisme.g., single non-transferable vote (SNTV) systems in Japan and Taiwan)encourage clientelism and vote-buying (Gallagher 1998;Persson et al. 2003), which then tilt the playing field towards the ruling party (Scheiner 2006;Wang and Kurzman 2007).
Against this background, Table 2 demonstrates that a predominant party system may emerge under wide-ranging versions of electoral systems, including majoritarian (India and Mexico) and SNTV systems (Japan and Taiwan). It has, though, been shown that nonproportional systemsplurality voting, block voting and SNTV systemsare more favourable for the reproduction of dominance than proportional systems (Dunleavy 2010: 13).
Among the cases with predominant party systems, the effect of the electoral system seems to range from major impact to none at all. For instance, in Taiwan and Japan, the dominant parties took great advantage of the SNTV system, to a much higher extent than in any other polity (Cox and Niou 1994: 221). 18 Notably, the electoral system awarded government to the LDP alone, although it had not succeeded in passing the 50 per cent threshold since the 1960s. Similarly, India's first-past-the-post (FPTP) system, which was adopted due to concerns for political stability within the country's highly diverse demographic structure, substantially empowered the INC against the fragmented opposition. 19 For instance, in the 1971 elections, the INC captured 350 seats (67.5 per cent) in the 518-seat Lok Sabha with a vote share of 43.6 per cent. In the same vein, in the 1957 elections, the INC won 361 out of 494 seats (73 per cent) with a vote share of 44.7 per cent. In the case of Mexico, on the other hand, the majoritarian system had little 16 The new institutionalism has centred on political institutions to understand variance across time and space. This is crucial because political institutions including the electoral system "structure the incentives that guide individual actions, which in turn underline the aggregate outcomes -such as the elections results and legislationthat political scientists observe." (Stoll 2013: 2). 17 Notable exceptions include Bogaards 2008;Diaz-Cayeros and Magaloni 2001;Reed 2007. 18 Due to the electoral system, intra-party factionalism boomed and partisan identification weakened in both cases (Grofman et al. 1999;Wang 1996). 19 This is in stark contrast to what Duverger (1959) envisaged: plurality rule in the single-member districts leads to two-partyism.
impact on the PRI's electoral success, as its vote share was far beyond the 50 per cent threshold. Among the aforementioned cases, only India still uses the same electoral system. In the remainder, the electoral system was changed after the opposition captured power (e.g., Japan and Taiwan) or as the incumbent desired (e.g., Mexico). Mexico 20 and Japan switched to mixed electoral systems in the mid-1990s, and Taiwan introduced a mixed electoral system in 2005. The impact of the electoral system change varied across cases. While the PRI and the KMT lost their majorities in less than a decade, the LDP kept its majority as late as 2009. This shows that the electoral system alone cannot fully account for the reproduction of dominance, and therefore needs to be considered along with other factors, such as the salience of social cleavages and the use of clientelist inducements.

The Electoral System and the Emergence of a Predominant Party System in Turkey
Turkey has seen many changes to its party system over the decades, making it a good lab for testing theories of predominant party systems. Empirical evidence demonstrates that it has two experiences with predominant party systems. The first one dates back to the 1950 election that marked the end of single-party rule and the victory of the Democrat Party (DP). The DP repeated its success in 1954 and 1957, thus transforming the party system into a predominant one. However, given the DP's rising authoritarianismevidenced in its efforts to shut down the Republican People's Party (CHP)by the late 1950s the party system had transformed into a hegemonic one.  1955-2009 (except 1993-1994) Taiwan (KMT) In the legislative Yuan the SNTV system was used until 1992. During the 1950s, Turkey operated a FPTP electoral system in multi-member districts. In contrast to what Duverger (1959) would predicti.e., plurality rules tend to produce two-partyism with centripetal tendenciesthe FPTP system did not lead to two-partyism in Turkey because of the absence of alternation. Neither did it lead the major parties to soften their political positioning and approach the centre. Although the CHP and the DP combined won over 90 per cent of the seats in every election, the two-party system, in theory, was nothing more than the maintenance of one-party rule. As Table 3 demonstrates, the margin between the vote share and the seat share of the parties was wide and the electoral system substantially over-represented the largest party. This shows the crucial role played by the electoral system in shaping the party system in the 1950s. The FPTP electoral system was replaced by the D'Hondt formula after the military intervention in 1960. This electoral engineering aimed to prevent the re-emergence of a hegemonic party system. Empirical evidence also supports such an expectation. Bormann and Golder's dataset (2013) shows that among 37 countries that use the D'Hondt system, most of them do not develop a predominant party system. This is mainly because the D'Hondt formula is designed to 'minimize over-representation of the most overrepresented party' (Gallagher 1991: 34).
What complicated the issue is that the D'Hondt formula is, at the same time, the least proportional of PR formulas (Lijphart 1990). The electoral system becomes even less proportional when the D'Hondt formula is combined with a high national threshold (10 per cent in Turkey), as this reduces the number of tiny splinter parties in the party system 21 (Farrell 1997) and increases the degree of disproportionality in favour of the larger parties (Anckar 1997). Table 3 sets out the discrepancies between the vote and seat shares of the parties in the elections from 1950 to 1999. The results show that the electoral system had a decisive impact on the parliamentary composition. For instance, in the 1987 elections, the Despite over-representation of the larger parties, the new electoral system failed to reduce fragmentation in the parliament. On the contrary, as Table 4 suggests, until 2002 fragmentation was rising: it jumped from 2.85 in the 1983 elections to a record-high 6.78 in the 1999 elections. This led to a series of coalition governments during the 1990s, unlike the one-party governments that characterized the 2000s.
How, then, can we explain how the same electoral system prompted polarized multipartyism during the 1990s but a predominant party system during the 2000s? We cannot address this question properly if we do not take into account the single-party governments of the ANAP during the 1980s. This means that the D'Hondt formula and a threshold, in the first instance, produced one-party governments. However, pervasive corruption and political inefficacy undermined their credibility and strengthened the opposition parties on both the left and the right of the political spectrum. This brought forth coalition governments: between 1991 and 2002, nine governments were formed, lasting 1.3 years on average. They were unstable and prone to political and economic crises, as in 1994 and 1999. In addition to low-performing governments, the lack of intra-party democracy and personal animosities between party leaders increased fragmentation in parliament. Similarly, party closures (Casal B ertoa and Bourne 2017; Tezc€ ur 2012) increased volatility and destabilized the party system. Taken together, the military's electoral engineering failed to produce the intended political outcomeat least with regard to reducing party fragmentation.
The second experience with a predominant party system dates back to the snap elections of November 2002. Wasted votes amounted to 47 per cent, as two parties on the right, the Volatility is based on the Pedersen formula. As the number gets closer to 0, parties' vote share remains the same across two elections. Fragmentation is calculated according to the Laakso-Taagepera effective number of parties' indicator (vote shares). If one party has a huge majority, the number is slightly above 1. Competition measures the vote-share differences between the winning party and the second-strongest party subtracted from 100. The nearer the resulting value to 100, the more competitive the party system. Disproportionality is based on the Gallagher index and is calculated by summing absolute differences between parties' seat and vote shares.
Does the Electoral System Foster a Predominant Party System? Evidence from Turkey 133 MHP and the DYP, fell only slightly below the threshold, receiving, respectively, 9.5 and 8.3 per cent of the vote. Table 4 demonstrates that turnout was exceptionally low in 2002 (76 percent) and then rose to above 80 percent as polarization grew. Moreover, after the AKP took office, the party system became less competitive, marked by the widening margin between the AKP and the second party (CHP). As an indication of a stabilizing party system, volatility dropped sharply from the highest point of 41.7 per cent in 2002 to 11.4 per cent in the June 2015 elections. This was because votes started to concentrate among four main parties. Similarly, fragmentation, which denotes the effective number of parties based on vote shares, decreased from 5.4 in 2002 to 2.9 in the June 2015 elections.
In addition, the level of disproportionality, indicating the difference between the vote and the seat shares of the parties, changed dramatically, mainly due to the electoral threshold. For example, when only two parties surpassed the threshold in the 2002 elections, the AKP secured almost two-thirds of the seats with a mere 34 per cent vote share. The disproportionality was 27.0 in the 2002 elections, the highest in the post-coup period. Disproportionality dropped sharply, however, to 4.9 in the November 2015 elections, as the number of parties achieving the electoral threshold doubled.
Disproportionality is more evident in the analysis of the provincial results from 2002 to 2015, especially in Kurdish-majority provinces (see Table 5). To illustrate, in the 2002 elections, the AKP won 6 out of 7 seats in Van with only 25.8 per cent of the votes. The same held true for the 2007 elections, albeit at a lower level. In 2011, disproportionality declined further, since the pro-Kurdish candidates ran as independents. In Batman, by way of illustration, the AKP won 3 deputies with 20.6 per cent of the votes in 2002, but in 2011 it managed to win only 2 seats with a 37.1 per cent vote share. Disproportionality dropped to record-low levels in the June 2015 elections. By comparison, in the 2002 elections, the AKP won eight out of ten seats in Diyarbakır province with a vote share of just 15 per cent. However, when the Peoples' Democratic Party (HDP) succeeded in passing the threshold in the June 2015 elections, the AKP won only one seat with the same vote share, whereas the HDP won ten.
In the empowerment of the AKP, district magnitude also played a crucial role, as anticipated by Rae (1967) and Lijphart (1994). 22 We must note first that the average district magnitude in Turkey was 6.25 (before the June 2018 legislative elections), which is considerably higher than in Japan and Taiwan under their SNTV electoral systems, but lower than the median among countries with PR rules (7.925) in Bormann and Golder's dataset. The same dataset shows that among 630 elections conducted under pure PR rules, only 196 of them (31 per cent) had average magnitudes smaller than Turkey's (leaving aside cases with plurality and majority rules, mixed systems and SNTV).
The election results demonstrate that the higher the district threshold (that is, the lower the size of the M), the stronger the AKP (see Table 6). The examples are abundant. For instance, in the 2002 elections, the AKP won 55 per cent of the seats in two-seat districts with, on average, a mere 25.2 per cent of the vote. It also won 58 per cent of the 11+ member districts with an average vote share of 33.2 per cent. The same trend is apparent in the 2007 and 2011 elections. Conversely, in June 2015, when both the MHP and the HDP passed the threshold, district magnitude became less salient in shaping the party system. To illustrate, the AKP won 53.1 per cent of the two-seat districts with an average vote share of 43 per cent. In the 11+ member districts, this margin between the vote (39.2 per cent) and the seat share (43.7 per cent) dropped as low as 4.5 per cent.
Thus far, we have attempted to show how the electoral system fosters the AKP's dominance. 23 However, one also needs to explain why the same electoral formula with the same threshold produced fragmentation during the 1990s, but consolidation after 2002. As far as we can discern, the divergence has both electoral and non-electoral causes. First, the dynamics that laid the ground for the transition to multipartyism during the 1990s, namely inefficient governments and pervasive corruption, led to one-party governments during the AKP period, as the voters equated one-party government with stability. Equally important, there were some elements that stemmed from the electoral system itself.
To start with, turnout remained very high (Table 5). This might reflect two contradictory tendencies. The first is that voters might internalize the rules of the game. In this sense, turnout serves to legitimize the political system. The second is just the opposite. The electoral system de-legitimizes the political system by fostering polarization, which motivates voters to go to the polls. In this sense, high turnout levels merely express the level of polarization in society. 24 It is our view that the latter overwhelmed the former, especially in recent elections. Notes: *Bayburt elected one deputy in 2011. 23 An overview of the literature reveals that the role of the electoral system on the emergence of the predominant party system has not been systemically discussed. Studies (e.g., C ß ınar 2016; Sayarı 2016; Tezc€ ur 2012) briefly emphasize the role of a 10 per cent nationwide threshold in parliamentary composition, but fail to show how the electoral system or district magnitude affects the provincial results. In addition, with the superficial exception of Ayan-Musil's (2015) short discussion, the existing literature lacks any theoretical framework and reliable findings that would situate Turkey in a cross-national perspective. 24 For an excellent study of the dynamics and the current situation of mounting polarization, see Esmer (2019). Relatedly, the electoral system, especially the electoral threshold, leads to a decline in wasted votes in the longer run (see Table 7), which coincides with the crystallization of the votes around major parties. This trend illustrates the "psychological effect" (Duverger 1959) in voting behaviour, although, to our knowledge, this argument has not so far been tested with individual data.
Additionally, the electoral system leads to strategic voting. Although it is true that all electoral systems allow strategic voting, 25 its forms and proportions vary. For instance, unlike in Europe, where strategic voting is used against "anti-systemic" parties such as the Front National in France or the Alternative for Germany (AfD) in Germany, in Turkey it is used against the incumbent. This is arguably the reason the HDP met the threshold in the June 2015 elections. The possible scenario was that, had it failed to do so, then the AKP would have formed government alone, possibly with a large enough majority to change the constitution unilaterally. With this in mind, CHP voters in metropolitan areas switched to the HDP (Canyas ß et al. 2016;Grigoriadis 2016;€ Onis ß 2016). 26 The tendency towards strategic voting manifests the extent of polarization in society. The underlying assumption is that voters cast their votes strategically to avoid a wasted vote empowering the other camp. This level of polarization is also manifested in the troubling rise of hostile partisanship.
Moreover, the threshold not only worked as a barrier to new parties entering parliament, but also led to the decline of formerly mainstream parties. As an illustration, although numerous new parties were established across a wide-ranging ideological spectrum, and some of them, in fact, garnered enough power to potentially put the AKP's parliamentary majority in jeopardy (e.g. People's Voice Party), since 2002 none has acquired sufficient votes to enter parliament. Even more strikingly, mainstream parties of the 1990s -ANAP, DYP and DSP (Democratic Left Party)turned into marginal parties. This trend ultimately narrows the scope of political competition and reduces the quality of democracy.
Aware of the institutional barriers, the small parties in Turkey have followed two main strategies. One is aligning with other parties, large or small, before the election. A typical example is the alliance between the Nationalist Work Party (MC ß P) (then MHP), Welfare Party (RP) and the Reformist Democracy Party (IDP) in the 1991 legislative elections.  25 For instance, because the UK electoral system means that voters are choosing governments rather than parties, they are prone to strategic voting (Mair 2009). 26 Whether this sufficed for the HDP to pass the threshold is doubtful. Konda (2015), for instance, found that it was mainly due to AKP voters switching to the HDP.
Those minor parties who do not opt for this strategy are marginalized (e.g., Patriotic Party) or dissolve themselves (e.g., Anatolian Party). This is mainly because parties that are not represented in parliament are not eligible for treasury grants, so if there is no strong solidarity among supporters or private resources for funding party activities, they are destined to disappear. The second strategy, followed particularly by pro-Kurdish parties/candidates, is to run as independent candidates. Notably, in the 2007 elections, the pro-Kurdish party, Democratic Society Party (DTP), believing that it was unlikely to pass the threshold, ran with independent candidates and won 21 seats. The same strategy awarded its successor, Peace and Democracy Party (BDP) 36 seats in the ensuing elections. The HDP opted for this strategy out of necessity, to establish a group in parliament and benefit from treasury grants, although it would have won many more deputies if it had run under a party list.

Conclusion
This paper sought to address how the electoral system fostered the AKP's dominant party rule in Turkey. We first clarified the widely used but weakly operationalized concept of the predominant party system. Using a conceptual framework adopted from Sartori, we argued that to transform any party system into a predominant one, a predominant party must win at least three consecutive elections with more than a 10 per cent margin, and set up a one-party government in a competitive political environment. Having established the conceptual framework, we briefly investigated the explanatory power of diverse theoretical frameworks, observing that social cleavage theory and resource theory perform better in explaining the AKP's electoral success than performance legitimacy theory and decisiontheoretic models.
We highlighted how the electoral systemthe D'Hondt formula with a 10 per cent thresholdled to the predominant party system in Turkey. Our analysis revealed a threefold causal link between the electoral system and the predominant party system. First, the electoral system over-represents large parties and under-represents small ones. This is because the D'Hondt formula (the least proportional among PR formulas) and a 10 per cent threshold (the highest in the world) combine to hinder the election of new parties, to weaken the old ones (e.g., DYP and ANAP) and to motivate small parties to align with others or run as independents. As demonstrated, disproportionality was highest in Kurdish-majority provinces. Second, the electoral threshold prompted strategic voting: that is, it induced voters to opt for their second-best option. This concentrated votes among the larger parties and minimized wasted votes. Third, the electoral system boosted the turnout ratefrom 76 per cent in 2002 elections (the lowest in the post-1980 period) to 84 per cent in November 2015 elections (the highest since 1987)as it increased polarization and motivated voters to go to the polls.
In addition to the electoral formula and the threshold, the district magnitude also empowered the AKP. Specifically, we showed that the margin between the AKP's vote and the seat share was much wider in two-member districts than in 11+ member districts.
We also aim in this paper to situate the Turkish case among the broader set of predominant party systems, namely Mexico, Japan, Taiwan and India, on which the literature provides little or no evidence. Our analysis revealed that the electoral system has no systematic effect on dominance. For instance, the electoral system allowed the LDP in Japan to form government alone despite a vote share below 50 per cent over many Does the Electoral System Foster a Predominant Party System? Evidence from Turkey 137 decades, whereas in Mexico it had little impact, as the PRI's vote share was sufficient to manufacture parliamentary supermajorities. All in all, the electoral system in Turkey undermines democratic representation. What, therefore, is the rationale of politicians in maintaining such an electoral system, especially one with the highest threshold in the world? The first thing to note is that the maintenance of the present system serves the interests of governing parties in excluding their rivals from government. Second, this policy was designed to exclude Kurds from the parliament and decrease the salience of the Kurdish issuein fact, however, it has worked to the contrary, and intensified the armed conflict.
Recently, though, two challenges to this status quo have begun to develop. First, strategic voting against the incumbent among opposition voters, especially CHP supporters, worked well to break the barriers to electing the pro-Kurdish party in November 2015 and June 2018. Second, with the transition to a 'Turkish-style' presidential system, the electoral threshold has become a less effective barrier to small parties, as they are now allowed to form alliances before the elections and the aggregated vote share of an alliance counts towards the threshold.
Several avenues of research remain to be explored. This paper has highlighted the causal link between institutional elements and the predominant party system; meanwhile, more systematic research needs to be conducted to understand the role of neoliberal populism and economic performance in the AKP's dominance. Likewise, to develop a comparative perspective, new evidence should be sought to discover whether, and to what extent, district magnitude and strategic voting matter in other predominant party systems.