Direct democracy and its integrity. The Italian 2020 and the Turkey 2017 constitutional referendums

What is the role of referendums in different regimes from authoritarian electoral democracies to democratic systems? In this article we use the Direct Democracy Integrity Index to analyse the Turkish (2017) and Italian (2020) referendums to understand if they can be seen as contributing to majoritarianism or to consensus democracy, depending on rules such as who can call a referendum and on the integrity of the referendum itself and considering the misuse of the instrument by the executive especially in the initial phase of the referendum. Turkey and Italy are used as a most different systems design: a modern autocracy and a consolidated democracy are actually similar in their extensive and frequent use of nationwide constitutional referendums. We conclude that the constitutional referendum in modern autocratic Turkey contributed to majoritarian developments and strengthened the president, while in democratic Italy the referendum also offered incentives for minorities and regions to express dissent. Integrity was more problematic in the pre-referendum phase.

Representative democracy has been under intense scrutiny in the last two decades for its shortcomings. The models and practices of direct and deliberative democracy were seen as complements that could improve the quality of decision-making. In spite of extensive research, several details remain unclear: how elections, referendums or deliberative practices can include the under-represented groups; how many citizens and who can get involved in decisions; what form and aims of representative, direct and deliberative democracy can ensure the effective presence of underrepresented groups in decisions; what are the appropriate settings to ensure an inclusive engagement of citizens; and how can dialogue and reason accommodate the emotionality of participants and finally, what is the role of referendums in different regimes from authoritarian electoral democracies to democratic systems?
After the "unpleasant" results on regional integration (such as the Brexit referendum), referendums are highly criticized. According to the model of the democratic rhombus (Kersting 2014(Kersting , 2016, political participation is divided into four different political spheres: participation in representative democracy, participation in direct democratic instruments, deliberative participation, and demonstrative participation.
Here we focus on referendums as instrument of direct democracy accepted by most citizens as well as politicians.
In this article, we analyse the use of referendums in two Mediterranean countries, Italy and Turkey, to decide on policy matters and constitutional amendments, either through an "accept-or-refuse" question or as an optional alternative. In our article, we ask with Lijphart (1999) whether they can be seen as contributing to majoritarianism or to consensus democracy, depending on rules such as who can call a referendum and on the integrity of the referendum itself and considering the misuse 1 of the instrument by the executive dominance especially in the initial phase of the referendum.
In what follows, we will first justify case selection and describe data collection, we then turn to the description of the two cases chosen for the analysis. In paragraph 3, we show the results of the expert surveys while in the conclusions we discuss the integrity of the two referendums to illustrate similarities and differences.

Case selection and data collection
Given the focus of the Special Issue on Mediterranean Countries, we decided to study the cases of Turkey and Italy as a most different systems design. The democratic system in Italy and the electoral modern autocracy in Turkey differ extremely, but they are the two cases with the most frequent use of nationwide referendums in the Mediterranean area (5 constitutional referendums each). Despite the considerable use of referendums in both cases, Turkey can be considered a deconsolidating democracy, while Italy is not.
Referendums have direct and strong impacts on constitutional reform and policy change (see Altman et al. 2014;Ginsbourg and Dixon 2011). In Turkey, as an optional alternative for constitutional change, all referendums under the Adalet ve Kalkınma Partisi (AKP) government were accepted, while in Italy the 2016 referendum failed and the 2020 was approved. Analysing the integrity of the 2017 Turkish referendum and of the 2020 Italian referendum, we claim that the constitutional referendum in Turkey contributed to majoritarian developments, while in Italy the different variants also offer incentives for minorities and regions to express dissent.
In this paper, we use an empirical instrument, the Perceptions of Direct Democracy Integrity (DDI) index, to evaluate and compare the variety and integrity of the last Turkish and Italian referendums. The DDI, like the Perceptions of Electoral Integrity (PEI) index, has shown to be useful and relevant in measuring the integrity of referendums (Kersting and Grömping 2021).
Since 2012, the Electoral Integrity project (EIP) has analysed all parliamentary and presidential elections globally by means of the PEI Index (Norris 2014). The DDI index is instead a new instrument built to specifically analyse referendums quality. Like the PEI, the Direct Democracy Integrity Index draws on experts' assessments. An election expert is defined as a political scientist (or other social scientist in a related discipline) who has demonstrated knowledge of the electoral process in a particular country (such as through scholarly publications, membership of a relevant research group or network or political science association, or university employment or through snowballing). Data therefore come from an online expert survey that has been administered to Turkey and Italy experts after the referendums in question. The survey questionnaire includes fifty-nine measures of referendum integrity covering the whole referendum cycle cited above (see Appendix in Kersting and Grömping 2021). These items are grouped into eleven sequential sub-dimen- Fig. 1 The referendum cycle. (Source: Kersting and Grömping 2021) sions that reflect the dimensions of the referendum cycle (see Fig. 1). Experts are asked on a five-point Likert scale whether they agree with the statements (strongly disagree-to strongly agree) in around 60 integrity items. Survey invitations were sent to 229 Turkey experts in mid-June 2017 and to 436 Italy experts in May 2021, with two follow-up reminders one week apart each. Drawing on the work and on the theoretical framework of the Electoral Integrity Project Norris and Grömping 2019), in the following we compare the integrity of the Italian and Turkish referendums. In both countries, scholars and senior staff of political science departments as well as advanced political scientists who work abroad and who participated to international conferences were approached. Here the focus was on scholars working on internal politics or comparative politics with a focus on Italy or Turkey. The results were checked and controlled for inconsistencies and biased answers in the pro and contra groups. We received 45 responses from Turkey experts and 45 from Italy experts, representing a response rate of around 20% in Turkey and 10% in Italy.
First, we apply Direct Democracy Integrity Index on the referendum cycle model (see Fig. 1) based on the electoral cycle framework (European Commission 2006;Norris 2013). Like elections, referendums go through several phases and referendum quality can have shortcomings in electoral laws and electoral procedures as well as in voter registration, initiation of referendums, campaign and media coverage or in campaign financing. Problems may arise also in the process itself, voting, the post referendum vote count, and the role of the electoral authorities (Bedock 2017). This shows the necessity of an approach looking at the whole process because every phase has important areas for evaluation.
By comparing Italy and Turkey we expect that the modern authoritarian regime in Turkey and its president Erdogan manipulated and misused the referendum, while Italy, being a case of consolidated democracy, is expected to perform well on electoral integrity, even though some relative flaws may be found only in some specific K and more questionable issues, such as campaigning, that may lead to variation in experts' responses. In analysing the integrity of the 2017 and 2020 referendums, we will adopt the cyclical approach recognizing that procedural integrity can be broken at any step of the referendum cycle ( Fig. 1).
In the following, we thus present the results of the expert surveys conducted on the 2017 Turkish and the 2020 Italian constitutional referendum following the 4 phases of the cycle: pre-referendum, campaign, polling day, post-referendum. Direct democracy allows the electorate a direct vote on a specific political issue (see Qvortrup 2018). Direct democracy includes different types, legal settings and local and national contexts of referendums. According to who has the power of calling a referendum and whether the results are binding or not, we can have different types of referendums. Here we analyse Constitutional referendums, which imply a direct vote on a specific constitutional issue initiated by the Executive (president) or the legislative (parliament). Results are always binding.

Referendums in Italy and Turkey
Italy has a long history of referendums. Since the post-World War II, Italian citizens had the chance to vote in 74 referendums, mainly abrogative (Altman et al. 2014). According to the typology proposed by Kersting and Grömping (2021), Italian referendums can be classified into three main types: 67 abrogative referendums with bottom-up initiative; 5 constitutional referendums, which however in Italy are not mandatory and can have top-down or bottom-up initiators; and one consultative referendum.
Among the 67 abrogative referendums, 63 had popular initiative, while 4 were proposed by regional Councils. Among the 5 constitutional referendums, one, in 1946, was devoted to the choice of the form of the state (monarchy vs republic), while 4 were initiated by MPs (the 2006 referendum by MPs, regional Councils and citizens). The only consultative referendum was held in 1989 to probe the popular will regarding the conferment of a hypothetical constituent mandate to the European parliament 2 (Table 1).
Since the 1990s there have been 53 abrogative and 4 constitutional referendums. Among the 4 constitutional referendums, which do not require a quorum, 2 ap- proved the proposed reforms and 2 rejected them. Instead, among the 53 abrogative referendums, 28 did not reach the quorum. Electoral laws were the subject of referendums nine times. First, in 1991, citizens abolished multiple preferences for the election of the Lower House, thus reducing the available preferences from 4/5 (according to district magnitude) to one. In 1993, citizens abolished the qualified majority of 65% of the votes for the election of Senators in single-member districts, thus transforming the system in a first-past-the-post plurality. However, the result of this referendum was never applied since the 1993 electoral law, the so-called Mattarella Law (from the name of its rapporteur, the now President of the Republic Sergio Mattarella), introduced a mixed-member system. That law was again the subject of two referendums, in 1999 and 2000, to abolish the proportional quota of the mixed system at the Lower House; both referendums, however, did not reach the 50% quorum. Meanwhile, other two referendums touched upon election-related issues: in 1995, citizens did not repeal the rule that provides for a two-round system for the direct election of the mayor of municipalities with a population greater than 15,000 inhabitants and, in 2000, the referendum on the abolition of reimbursement to parties of electoral and referendum expenses did not reach the quorum. The same happened in 2009 to the last three referendums on electoral issues: the abolition of the possibility of multi-candidatures and the assignment of the majority premium to the most voted list, rather than to the most voted coalition, both in the Lower and in the Upper House (Regalia 2015).
The Italian referendum of 20th and 21st September 2020 was a constitutional referendum and asked citizens about their willingness to reduce the number of parliamentary seats: from 630 to 400 in the Lower House, from 315 to 200 in the Upper House 3 . The discussion about the reduction of the number of MPs dates to the 1980s and was the object, among other things, of two previous referendums, one in 2006 and one in 2016, which however were outvoted by a majority favourable to the status quo. The 2020 reform originated from the government agreement signed by the 5 Stars Movement (M5S) and the Lega in 2018 and was took forward one year later by the subsequent government coalition by the M5S, the Democratic Party (PD) and the left (LEU). Having been sponsored by two different government majorities, the reform was not very controversial. In fact, over two thirds of the Members of the Lower Chamber approved the text. However, Art. 138 of the Italian Constitution provides for the possibility of requesting a constitutional referendum if the two Houses approved, after the second vote, a constitutional revision law or a constitutional law by absolute majority. It will not be possible to request a referendum if a qualified majority of 2/3 of the members of each House is reached in both Houses. The request can be presented by one fifth of the members of a Chamber, by 500,000 voters or by five regional Councils within three months of publication in the Official Gazette. Only the Chamber of Deputies approved the constitutional reform with a 2/3 majority, and so one fifth of Senators (precisely 71 Members of the Senate), coming from different political parties (from left to right), asked for a constitutional referendum (Uleri 2021). The Conte II government (2019-2021; M5S and PD) campaigned in favour of the reduction in the number of MPs, while there was no strong party really against the proposed reform. Actually, among the main parties, only Berlusconi and Renzi's parties (Forza Italia and Italia Viva respectively) have shown some divisions among their ranks without declaring explicitly the positions of the two parties and thus allowing a complete freedom to their voters. No outreach programs were foreseen or enacted, and the referendum debate on the main television and printed media suffered from the competing attention given to the pandemic, government strategies in dealing with it, and its consequences. Unlike the abrogative referendum, the constitutional referendum does not provide for the achievement of an electoral quorum equal to at least 50% + 1 of those entitled to vote. With a voter turnout of 51.1%, the Yes vote won with a large margin: 70% of voters voted in favour of the constitutional amendments, and 30% voted against it.
The Turkish 2017 case was different in many respects. While not a fully authoritarian regime, Turkey is an example of a hybrid-regime, or at most of a semi-authoritarian one (Levitsky and Way 2002). As many studies have shown (Qvortrup et al. 2020), also non-democratic political systems resort to referendums. There, however, referendums (and elections) are often used to strengthen the position and the legitimacy of their authoritarian leaders (Kersting 2010(Kersting , 2011. Actually, numerous semi-authoritarian systems and mixed regimes have developed since 1989. Different labels try to characterize these hybrid authoritarian regimes such as modern authoritarian regimes, virtual democracies, illiberal democracies, defect democracies, electoral democracies, competitive authoritarianism, deliberative authoritarianism, etc. (Schedler 2006;Puddington 2017;VDem 2021;He and Warren 2011).
Here modern authoritarian leaders try to legitimize their domination by using democratic institutions such as elections and referendums. However, doing this includes the risk of losing the election or the referendum. To avoid defeat, authoritarian leaders try to control the outcomes through a variety of systems including control of the media, propaganda, and a comprehensive control of public opinion. On the other hand, constitutional referendums in democratic system are increasingly including deliberative instruments (Dryzek 2002;Welp andSoto 2020, Kersting 2023). Also, illiberal non-democratic regimes use additional forms of authoritarian deliberation in constitutional reform processes (He and Warren 2011).
The history of Turkish constitutional referendums dates back to the end of World War I (see Table 2 for turnout and result data) and sees a number of constitutions developing after military coups (Özbudun and Gençkaya 2009; Özerdem and Whit- ing 2019). After the short-lived constitution, which was implemented immediately after World War I, in 1921 a constituent assembly drafted a new constitution, which was adopted by parliament in 1924.
In 1961, after the 1960 coup, a new constitution entered into force. The constitutional process was dominated by the organizers of the coup, and it was approved with the referendum in 1961. After the 1980 coup, the military dominated a new constituent assembly, which led to the Constitution of 1982 that was approved by referendum (Özbudun 2015). In the following years, the ruling parties tried to change the Constitution and the electoral law which led to various confrontations. Because the proposed changes did not get the two-thirds majority in parliament in 2007 and 2010, the amendments had to be legitimized by two referendums. From 2011 to 2013, the broader constitutional reform process was set in place allowing outreach programs, legal participation of political parties as well as civil society (Adar and Seufert 2021). The submission of individual proposals to the constitutional conciliation commission was made public through the Constitution making website and citizen consultations were held across the country. This mechanism tried to incorporate civil society and triggered the broad process of citizen deliberations. In this process, a broad consensus was reached on 16 articles including 43 chapters focusing on freedoms and fundamental rights. Nevertheless, after 18 months the whole process failed due to the ruling party insistence on implementing the pure presidential system (Adar and Seufert 2021).
After the failed military coup in July 2016, a state of emergency was declared, and a new process of constitutional reform began. In the pre-referendum phase, the debates of the constitutional amendments were highly controversial and the situation was tense. This and the post-coup measures led to the dismissal of tens of thousands of public sector officials, teachers, professors etc. and to an atmosphere of intimidation. In this state of emergency context, the AKP in coalition with the nationalist Milliyetçi Hareket Partisi (MHP) proposed its controversial constitutional amendments in late 2016. Because this alliance could not get a 2/3 majority in parliament, a referendum became necessary (Ekim and Kirisci 2017). The campaign phase felt in a time of political repression and intimidation. A substantial amount of the population migrated to other European countries. On the other hand, the established Turkish diaspora was regarded as conservative. Most of these migrants were allowed to vote, and about 1.4 million cast their ballot from overseas.
In Turkey, the political situation and the context of the analysed 2017 referendum was different from the constitutional review and the referendums beforehand. The referendum was held in highly tense and polarized circumstances. This confrontational context followed a military uprising after an attempted coup on July 15, 2016 (Kalaycıoglu and Kocapınar 2021, p. 406).
The legal framework provides for a mandatory referendum at the national level (see IDEA 2008; Qvortrup 2018). There are no other direct democratic instruments at the subnational level. Since there are no recall and agenda setting referendums, referendums can only take place on constitutional changes. If a two-thirds majority in parliament does not approve constitutional amendments, a referendum must be held. Furthermore, the Turkish constitution lists legal provisions for optional referendum, while it does not foresee the possibility for agenda initiatives at the national level. Results of national referendums are always binding.
Altogether, about 46 million people voted in the popular referendum held in April 2017 in Turkey and abroad, a turnout of 85.3%. However, the yes-vote won only with a very small margin: 51.4% voted in favour of the constitutional amendments, and 48.6% voted against it (Zirh et al. 2020). In Germany, the Turkish foreigners abroad gathered around 660,000 votes. Here a clear majority of 63% voted in favour of the constitutional amendments and 37% against it (Makovsky 2017). In Turkey itself, there was a substantial split between urban and rural areas. In particular, the areas close to the Mediterranean Sea as well as in the East of the country (with the Kurdish population) voted against the constitutional amendments. Most of the rural population in central Turkey-with the exclusion of the metropolitan area of Ankara-and in the North of Turkey voted in favour (on the urban/rural cleavage in Turkish vote, see Çarkoglu and Kalaycıoglu 2021).
As a result of the binding constitutional referendum, the 2017 constitutional reform introduced a strong presidential system (see Adar and Seufert 2021;Kersting and Grömping 2021). The approved amendments to the constitution abolished the office of the Prime Minister and strengthened the executive role of the President in the Turkish political system (Öney 2018). The change from a parliamentary system towards a presidential system included the considerable expansion of presidential rights. The Turkish president can now appoint the Supreme Board of Judges and Prosecutors (Hâkimler ve Savcılar Yüksek Kurulu, HSYK). While a small number of provisions in the referendum package strengthened the parliament (the number of seats in parliament was also increased from 550 to 600), in general, power was centralised, and checks and balances were weakened (Yılmaz 2020).

The referendum cycle and its integrity
In this paragraph, we analyse the results of the two expert surveys conducted after the 2017 and 2020 referendums. Figure 2 shows the eleven dimensions of the referendum cycle, plotting for each dimension the average percentage of agreement with survey questions denoting higher integrity. The left panel refers to the Italian 2020 referendum while the right panel to the Turkish 2017 referendum. The comparison between the two countries is obviously stark: Italy is a case of consolidated democracy while Turkey shows the nature of its authoritarian regime.
The graph shows that, for Italy, the most problematic dimensions were campaign, media coverage, campaign finance and post-referendum environment, while the process was in general considered of high integrity. On the contrary, the referendum process in Turkey was deeply criticized also by electoral observer groups such as the one from the Organisation for Security and Co-operation in Europe (OSCE). The observers complained about the lack of a level playing field before the referendum. While, according to OSCE observers, the technical administration of the referendum was relatively well managed, the access to the media was deeply criticised as uneven, abundant negative campaigning against the 'No' side took place (denouncements as "terrorist"), adequate resources for effective redress of electoral board decisions were lacking, among other things (OSCE 2017). The criticisms were so harsh that Erdogan government accused these observers of being politically biased.

Pre-referendum and initiation
The legal framework is considered fair by two third of experts in Italy and only 12% in Turkey. It is interesting to confront the answers regarding the agenda setting: although the Turkish referendum was mandatory because of the lack of the 2/3 majority in parliament, while the Italian referendum was not, more than 90% of the experts in Turkey and only 45% in Italy stated that the executive was dominating the agenda setting for the referendum (see Fig. 3 below). Opinions regarding whether the referendum law favoured the status quo side are almost opposed in the two countries: while only about half of the experts (45%) stated that referendum laws did not favour the status quo side in Italy (27% disagreed), 61% (vs 27%) believes the law favoured the status quo in Turkey. Voter registration in Italy is automatic, Fig. 3 Pre-referendum and initiation. Integrity in Italy and Turkey. Direct Democracy Integrity Index: 5 Likert scale ((Strongly) disagree-(strongly) agree). Middle category ('neither agree nor disagree') not shown but can be deduced K so almost all respondents deemed the electoral register accurate (without ineligible citizens registered or eligible citizens not registered). On the contrary, 46% of Turkey experts believes it was inaccurate, underlining again the differences in the integrity of the pre-referendum and initiation phases in the two countries.
In general, 77% (vs 2%) of respondents sees the whole initiation process in Italy as legitimate and fair, while in Turkey only 9% (vs 89%) considered the initiation of the Turkish referendum legitimate and fair. This is easily explained by the full democratic context in which the Italian referendum took place and by the hostile context and the state of emergency after the coup attempt in 2016 that affected the Turkish referendum.

Campaign
As mentioned, campaign was the most problematic phase in Italy (see Fig. 4). Differently from Turkey, although the 2020 question was not a difficult one because it simply asked citizens to approve or disapprove the reduction of parliamentary seats (58% of the experts (vs 24%) deemed the question on the ballot unambiguous), the implication of this choice for the democratic functioning were not straightforward.
Thus, the campaign in Italy had to be balanced to provide accurate, accessible, and relevant information. Most respondents believe citizens had sufficient time (59% vs 18%) and information (53% vs 22%) to discuss and understand the issue and no group was restricted from holding campaign rallies (71 vs 2%), while in Turkey 86% (vs 9%) of respondents asserts there were restrictions.
Media coverage in Italy was seen as problematic, although both sides had equitable access to the media to broadcast their messages (51% vs 20%). In Turkey, on the contrary, only 9% (vs 91%) of experts believes that both sides had fair access to political broadcast and advertising. The regulation and practice of campaign finance is considered almost fair in Italy but extremely unfair in Turkey: most respondents believe that no state resources were improperly used (39% vs 12%) in Italy, while this percentage is only 5% (vs 93%) in Turkey. In Turkey, 95% deny that both sides had equitable access to public political subsidies, 84% say that both sides had equitable access to political donations and 93% denies that both sides published transparent financial accounts with only 2% disagreeing.

Polling day
In Italy, in terms of referendum procedures, almost unanimously respondents consider the referendum well managed (67% vs 12%), conducted in accordance with the law (88% vs 4%), with fair election officials (84% vs 4%) and with an adequate number of polling stations (82 vs 4%), as expected in a consolidated democracy. Also, information about voting procedures were widely available (71% vs 8%) (see Fig. 5). This was not true in Turkey, where only 47% of experts (vs 40%), considers information widely available.
The voting process in Italy was without any doubt considered free and fair and voters did not fear becoming victims of political violence (76% vs 4%). On the contrary, in Turkey political violence was widespread (19% vs 69% of respondents).
In Italy, more than two-thirds of respondents, moreover, believes that no fraudulent votes were cast (69% vs 2%), while in Turkey 86% (vs 10%) of the expert criticized this practice. More than three quarter of the experts also mentioned that some voters were threatened with violence at the polling stations. The same amount said that people were not free to vote and there was feeling of pressure. The vote count, one of the most sensitive phases, in Italy it was evaluated fair (84%). This was highly problematic in Turkey, where 79% complaint about the lack of integrity.

Post-referendum
In Italy, the post referendum phase was not very problematic and results were announced promptly (69% vs 18%). Very few experts think that parties or individual challenged the results (6% vs 78%) (see Fig. 6). On the contrary, in Turkey 98% of the experts confirm result challenging: actually, complaints from different opposition groups demanded the annulment of the referendum because many votes were expressed on non-official ballot papers. The decision of the higher electoral commission which allowed non-official ballot papers to be accepted was deeply criticized. The observer groups of the Council of Europe and OSCE also complained about this decision but could not verify any cases of manipulation and electoral fraud in this regard. Pro-Kurdish political party Halkların Demokratik Partisi (HDP) announced a complaint to the European Human Rights Court. In Turkey, the post-referendum phase continued under state of emergency legislation.
Thus, it does not come as a surprise that 86% of the experts do not see the electoral authorities as impartial (12% disagrees). 81% of the experts stated that the authorities did not perform well. On the contrary, in Italy, the electoral authorities administering the referendum were seen as impartial (69% vs 6%), performed well (69% vs 6%), distributed adequate information to citizens (59% vs 10%) and allowed public scrutiny of their performances (45% vs 8%).
In Italy, experts agree that the post-referendum environment was not violent (84%) and did not give rise to protests (69% vs 8%). Disputes were solved though legal channels, say 24% of the experts (12% disagree), and voting results were subject to a postelection audit (41% vs 8%).

Discussion
Referendums are used in democratic system as well as in modern authoritarian regimes. Like elections, referendum integrity can be very different. In general, it was expected that the established Italian democratic system would get much better scores by the experts then the authoritarian regime in Turkey. In the following, we discuss and compare the four different referendum phases to understand if the integrity of the two processes has similarities.
In the case of direct democracy cycles, the pre-referendum phase and the initiation-phase are very important. In these phases, it is relevant whether there is a possibility to initiate referendums through a bottom-up process, as in Italy, or if referendums are mandatory, as in the case of the Turkish constitutional referendums. The strength of the president is related to his/her right to call for a referendum whenever s/he feels this could reinforce his/her own position. For this reason, referendums were often labelled a "bonus for dictators". In the Turkish case, the constitutional reform was tabled by the ruling party and its allies, knowing that the proposed reform would not get a two-thirds majority in parliament, and that the referendum would be mandatory. This shows that, in the end, the initiation of the referendum was like a plebiscite in the hands of President Erdogan, who did not expect a negative result. In contrast, Italy does not have a legal provision for mandatory constitutional referendums, but there are different bottom-up direct democracy processes (including popular initiatives) mostly related to abrogative referendums.
The Italy experts' evaluation showed that more than half of the experts evaluated very positively the different aspects of referendums laws, initiation, and voter registration. Nevertheless, compared to other phases, these first phases did not get full marks. Performances are criticized especially regarding initiation. At the same time, also the campaign phase was critiqued: experts believe that a broad public debate about the referendum did not take place and some experts regarded media coverage as well as campaign-finance as critical. In addition, a clear majority judged the Turkish legal framework and the initiation phase highly problematic. Similarly, campaigning was also very challenging: especially media coverage was regarded as an uneven playing field and campaign-finance was highly unfair in using governmental money for the interests of the ruling party. These deficits, therefore, were not only common comparing Italy and Turkey. If we look at the different phases of the Turkish referendum, during initiation and campaign, it was highly obvious that the integrity was also very critical.
When it comes to the polling day, it can be shown that the different procedures in the polling station as well as the vote count were getting slightly better marks in Turkey (compared to the other phases). The voting process was highly influenced by the state of emergency and by the context of intimidation. In Italy, quite obviously given its full democratic status, the procedures, the voting process at the vote count could fulfil the highest standards and got much better marks than the two previous phases.
The Italian post-referendum environment showed smaller conflicts in politics but not at all related to supposed malpractice during the referendum. This can also be deduced by the relatively good marks given to the Italian referendum authorities. In Turkey, if compared to the other phases, the electoral procedures and the vote count got better marks than the initiation phase. This is also the case for the post referendum environment. Here, the experts gave low marks since the Turkish referendum authorities are regarded as highly biased and unfair. Nevertheless, the score in the post-referendum phase was higher than in other areas.
We can conclude that, both in democracies and in electoral authoritarian regimes, voting processes and procedures as well as the vote count can still be problematic, as it was in Turkey, but they are much better organized and fair than most other phases of the direct democratic circle. The administrative capacity of the electoral institution did improve. It seems that open vote-rigging and electoral malpractice are becoming less important at the polling day. Much more problematic are the first and the second phase of the direct democratic process in democracies as well as in electoral authoritarian regimes. The initiation and the process of discussion before the referendum (during campaign) show numerous deficits and a lack of outreach programs also in full-fledged democratic systems. This shows that the Direct Democratic Integrity Index (DDI) is also a useful instrument in democracies. The lack of participatory instruments and public deliberation before constitutional referendums is also a liability of authoritarian regimes. Nevertheless, the debate in Turkey in the 2010s as well as deliberative outreach programs and other systems have shown that deliberation can also be implemented in authoritarian regimes. More problematic seems to be the dominant position of political leaders when it comes to media coverage. Here, in democratic systems as well as in autocracies, unequal campaign-finance and media monopolies privilege the ruling parties or other important interest groups, which control the initiation of referendums and the campaigning.

Conclusions
Comparing referendums in Italy and in Turkey means focusing on two "most different" political systems. After the authoritarian fascist regime and the Second World War, Italy became a fully-fledged democracy, which had-although legitimacy for the political system was low-numerous political reforms and constitutional amendments. These strengthened the democratic system. In Turkey, after the years of the military rule, the democratic system was implemented. This democratic system in the following years became dominated by the AKP and by President Erdogan. Turkey became more and more an electoral democracy with characteristic of modern authoritarian regimes. The referendum in 2017 did not follow the constitutional review process as in the early 2010s, with relatively broad constitutional debates and discussions. In contrast, the 2017 referendum was implemented in a state of emergency with the aim of strengthening presidential powers. In these two different Mediterranean cases, Italy and Turkey, referendums seem to play an important role for the legitimacy of the incumbents as well as for the articulation of interest groups in the civil society. However, the democratic meanings of the two referendums were really different: while in Italy the process remained in the boundaries of a democratic, con-solidated, polity, the Turkish referendum contributed to a presidentialisation outside democracy.
While referendums can be highly manipulated by authoritarian leaders, and despite all the executive misbehaviour and malpractice, the Turkish referendum showed a relatively close result, where the opposition got the majority in the metropolitan areas and lost only in the rural areas. Also, in the more touristic Aegean coastline an opposition stronghold becomes visible. This result can be considered as a trigger for the mobilization of the opposition. Actually, opposition leaders were very successful in the following local elections in bigger cities. For the authoritarian leader the referendum had positive and negative effects. The finally "successful" referendum introduced a new constitution, which strengthened the executive president position and weakened the role of the parliament. The strong president could in turn use these new powers to call for plebiscite to strengthen even more his position. Nevertheless, the unexpected very sharp result also seems to have weakened his power in the electoral authoritarian regime.
The analysis developed during this article brings us to some tentative conclusions. The constitutional referendum in modern autocratic Turkey contributed to majoritarian developments and strengthened the president, while in democratic Italy the referendum also offered incentives for minorities and regions to express dissent. This happened for two different but interrelated reasons. First, in Italy, the rules regarding referendums initiation provide minorities with a powerful instrument to raise their voices and block the decisions of the parliamentary majority, while in Turkey a strong majority can rule out such a possibility. Second, the referendum quality, especially in the first two phases (pre-referendum and campaigning), contributed to transform the referendum in an instrument of the executive in Turkey, and of the minorities in Italy. We can therefore conclude that initiation rules and electoral integrity are two powerful factors that can shape majoritarian or consensual developments both in democratic and in semi-authoritarian regimes.
Funding Open access funding provided by Università degli Studi di Milano within the CRUI-CARE Agreement.
Conflict of interest N. Kersting and M. Regalia declare that they have no competing interests.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4. 0/.