Change and Continuity in Turkey's June 2018 Elections.

• Author: Carkoglu, Ali
• Position: ARTICLE

Introduction

In arguably the most critical election in modern Turkish history, incumbent President Recep Tayyip Erdogan received nearly 26 million (52.5 percent) of about 50 million valid votes in the elections of June 24, 2018. Following a narrowly approved referendum in April 2017, President Erdogans new term starts with sweeping executive privileges that include the power to issue decrees with the force of law and to appoint the cabinet and vice-presidents as well as senior judges amongst other civil servants. However, in contrast to its repeated success in the second presidential election since 2014, Erdogans party, the Justice and Development Party (AK Party) lost its parliamentary majority. Although it remains the largest party with 295 seats in a parliament with 600 seats, the AK Party's electoral support declined from 47.8 percent in November 2015 to 42.6 percent in June 2018.

Muharrem Ince was the leading opposition candidate against Erdogan and received about 31 percent of the votes. Similar to Erdogans electoral performance compared to his party's in the parliamentary election, Ince's party, the Republican People's Party (CHP) did poorly and received only 22.7 percent of the votes, securing 146 seats in the parliament. This marks a noteworthy reduction of votes from 25.3 percent in November 2015. The left wing Kurdish Peoples' Democratic Party (HDP) once again remained above the 10 percent threshold and obtained 67 seats. HDP leader Selahattin Demirtas campaigned from behind bars and garnered 8.4 percent of the votes, effectively being the third largest candidate in the presidential election. The Nationalist Movement Party (MHP) supported Erdogan's candidacy and formed an electoral alliance with the AK Party, performing unexpectedly well by gaining 11.1 percent of the votes that brought 49 seats. As such, the MHP's performance is better compared to its rival the lyi Party (iYlP) founded by splinters from the MHP, which obtained slightly below 10 percent with 43 seats. The IYIP had been formed in 2017 after its leader Meral Aksener resigned from the MHP and ran in the June election within an alliance with the CHP.

Our objective in this article will be to present a detailed geographic analysis of the election results across districts. We will compare parliamentary election results in November 2015 and June 2018, as well as presidential election results in 2014 and June 2018. To do so, we collected election data based on the official Higher Council of Election results at the ballot-box level. Based on this data, we will explore the extent to which the 2018 elections compared with previous elections.

The next section describes the background to the 2018 elections. We present a framework to explain the reasons for holding a snap election which was initially scheduled for November 2019. We argue that short-term factors based primarily on economic concerns, combined with long-term ideological differences, were critical in the decision to hold snap elections in June 2018. We also mention several security concerns stemming from the coup attempt in 2016. The third section presents what we call mobilized ballot-boxes and shows patterns of electoral competition in the most recent elections. The fourth section presents a comparative geographical analysis of the most recent elections by referring to previous elections. In this section, we present six politically distinct clusters across districts and describe geographical party competition in Turkey, based on the 2018 elections, and the previous parliamentary elections of 2015 and the presidential elections of 2014. The fifth section concludes with expectations based on our findings and the newly established political system in Turkey.

Background to the June 2018 Elections

The June 2018 election results differed to a great extent from what had been predicted by most polls and punditry preceding the elections. According to expectations, the presidential election would go into a run-off with a small likelihood of Ince as the winner. To describe these expectations, we collected 22 publicly available results of opinion polls since January 2018 which mentioned at least one of the presidential candidates who eventually ran for election in June 2018. Before the elections, in early June, there were eight such polls and their average expected vote share for Erdogan was 50.6 percent and for Ince, it was 26.9 percent. Figure 1 below presents public opinion results from these 22 different polls. (1)

Results from these polls indicate that in June, two polls by Sonar and Gezici public opinion companies predicted Erdogan to be below the 50 percent threshold while six polls by MAK, Konda, Genar, Metropoll and another one done for Bloomberg predicted Erdogan to be above the threshold. Since most of these results (under simplifying assumptions) included the 50 percent within their confidence intervals, uncertainty continued until election night. Although Erdogan had a clear advantage in the first round, if there was a runoff between him and Ince, most pundits predicted a competitive election with Ince garnering additional support from other candidates. (2)

Looking at the development of candidate support levels across 22 publicly available opinion polls, we observe that Erdogan's win in the first round be-came a possibility only in the last few weeks before election day. Only two polling results before June predicted Erdogan to be above 50 percent (May average polling results show Erdogan at 45.6 percent). However, the average of all eight polls for Erdogan in June was 50.6 percent. Ince's level of support slowly rose and appears to have remained stable around 22-27 percent until June (May average polling results show ince at 24.4 percent). Only two polling results showed Ince above 30 percent in June, and the rest ranged between 20 to 29 percent with an overall June average equalling 27.2 percent. Hence the average of eight polls for both Erdogan and Ince were under-predictions for both candidates by (52.38-50.6) 1.78 and (30.79-27.2) 3.59 percentage points respectively.

For Demirtas's candidacy, there appears to be a consensus from early in May through the end of the campaign that he would garner about 10 percent of the votes. Only four of these 22 polling results showed Demirtas to be below nine percent, while nine polls showed him to be above 11 percent. Aksener's level of support appears to be between 18 to 24 percent before May in a total of five polling results (the average being 20.5 percent). From May to June we observe a steady decline in Aksener's level of support down to about 9 to 12 percent. The average level of support in May for Aksener was 16.5 percent, and the June average was 10.5 percent. However, even in June, five of the eight polling results showed her level of support ranging from 10.2 to 12.5 percent. Considering the fact that Demirtas obtained only 8.3 and Aksener 7.4 percent of the of the valid votes, these predictions were clearly misleading.

In short, the polling results underpredicted the performance of both Erdogan and Ince while overpredicting the vote shares for both Demirtas and Aksener. Felicity Party leader Karamollaoglus poor showing in the election results was no surprise since he appeared only in three polling results above the margin of error.

A similar perspective can be obtained about the level of electoral support for the major political parties from a total of 34 polling results that were made publicly available from January through June 2018. We again brought them all within a comparable range of predictions that lie between 0 and 100 percent. Figure 2 below shows the averages of these polls over time. We observe that while the predicted share of votes for the AK Party on average declined slightly from January to June, the share predicted for the opposition CHP rose by about three percentage points. While the decreased average AK Party share for June was still an overestimate within the typical margin of error, the overestimate for the CHP was off the mark by about 3.5 percentage points.

While the other smaller party average vote shares reached in June for the HDP and iYiP were quite close to their realized vote shares, the average prediction for the MHP was significantly lower than the party's realized level of support in the June election. All eight available polling results in June predicted that the MHP would remain below the 10 percent threshold and hence would get into the Turkish Grand National Assembly (TGNA) solely due to the alliance it had formed with the AK Party. The predictions of these eight polling results ranged between 4.6 to 9.3 percent. (3) Our analyses of the geographical distribution of election outcomes across districts may provide at least a partial explanation as to why this large error took place. As we will underline below, the AK Party and the MHP have typically garnered their electoral support from the same geographic constituencies. In June 2018, the MHP appears to have benefitted from vote flows primarily from the AK Party. Such vote switches took place within the constituencies of the two large alliance members and hence might not have been considered as an effective party switch by the voters. Nevertheless, given the quite comfortably large sample sizes, especially toward the end of the campaign period, such large differences between the predictions and realized vote share for one of the major parties indicates a deficiency in the sampling and/or other data collection technologies being used in this sector.

The absence of a level field during the campaign and expectations of possible electoral irregularities may partly explain the surprise Erdogans win caused in the first round. Given the above-summarized polling results, many among the electorate expected the presidential election to be decided in a run-off between Erdogan and Ince. In this regard, the concessions of the opposition...