INTRODUCTION
Türkiye went through a local election process in 2024. The governing AK Parti’s declining vote share was the most salient result of the election. As high inflation has been adversely affecting many occupational groups in the last two years, the role of harsh economic conditions on the election result was largely discussed in political circles and the media. Especially occupational groups with fixed income were highly dissatisfied. Industrial workers with a minimum wage and service sector workers with a fixed monthly fee were far from satisfied with the general course of the economy. This was no different for the middle class and the salariat, as high prices declined their purchasing power. Among others, some groups were particularly dissatisfied. Retirees, especially those living in metropolitan areas were uncomfortable with high living costs. Unemployed graduates of education faculties were disappointed, as they were expecting to be hired after their graduation. While landlords relying on their rental incomes were restless with the government’s rent increase limit, tenants were dissatisfied with high rental prices. Peasants, who were already adversely affected by the drought, found the government's agricultural incentives unsatisfying. Following the elections, the AK Parti government has acted responsively and introduced a new economic program to cope with this widespread dissatisfaction in the society. Several structural steps, including austerity measures in the public sector and cumulative multi-stage taxes, were taken. The monetary policy committee decided to keep one-week repo auction interest rate — the policy rate — constant at 50 percent (TCMB). As a result, macroeconomic indicators started to improve by the beginning of June 2024. Borrowing costs have started lowering as borrowing interest rates started to fall. On the other hand, the main opposition party, the CHP, has politicized widespread dissatisfaction in the society and sent class-related top-down signals to the electorate. After the election, the CHP has organized several meetings with titles, particularly emphasizing its target social classes. “Labor meeting” with industrial workers in Gebze, “wheat meeting” with agricultural workers in Tekirdağ, and “education meeting” with teachers in İstanbul were the first attempts in this regard. Apparently, the CHP is trying to consolidate its working-class base, which constitutes a large proportion of the electorate, for a potential snap election. This is no surprise, as upon the CHP’s increase in votes in the 2024 local elections, a rapid demand for a snap election emerged in the party. At the very beginning, the CHP’s newly elected chairman, Özgür Özel, has ended discussions and started a political détente process with President Recep Tayyip Erdoğan. Nevertheless, immediately after, Özel increased his criticism of harsh living conditions and hinted at the possibility of a snap election in the Gebze meeting. With recent parliamentary arithmetic, a snap election seems unlikely, as 360 votes are required for the Turkish Grand National Assembly (TBMM) to decide to hold one. Recently, the main opposition CHP has only 127 seats in the 594-seat (600) TBMM. Together with 57 the People’s Democratic Party (HDP), 34 the Good Party (İYİ Party), 20 the Felicity Party (SAADET), 15 the Democracy and Progress Party (DEVA) and another 23 seats from other opposition parties, 253 seats in total are not enough to call a snap election. Nonetheless, at the end of this process, high living costs resulting from high inflation, along with the CHP’s class-related signals, brought the social class phenomenon to the centre of Turkish politics for the first time after the Cold War period.
Contrary to its recent increasing importance, class voting is an understudied area of research in Turkish political science literature. It may be owing to its insignificant effect on political attitudes and behaviors or it is simply an academic ignorance. However, social class has long been a focus of great interest in political science research in the industrialized nations of the West. A large body of literature recognizes the importance of social class as a strong determinant of voting behavior. Today, various strands of evidence in the literature lend support to the idea that voters’ social class locations affect their voting decisions in the industrialized Western countries. When compared with Western countries, the class structure is suggested to be weak in Türkiye. It’s short capitalist history might have restrained a strong link to be established between social class structures and voting behaviors. On the other hand, the strength of non-economic cross-cutting factors such as religiosity, kinship, and fellow-townsmenship might have prevented social class from gaining a relevant role as a determinant of Turkish politics [1]. Nevertheless, post-1980 uninterrupted marketization and capitalization processes and new middle-class growth might have strengthened class based voting behavior. It is also likely that increasing ideological divergence might have exerted a supply-side effect and contributed to the strengthening of the class-voting link [2]. The fact that this gap stands in the middle of Turkish political science research and the recent strengthening link between social class and politics brought the long-delayed question to the research agenda: To what extent do Turkish voters’ occupational class locations affect their voting decisions? Based on above-mentioned assumptions, it is hypothesized that social class effect on Turkish voting behavior could be insignificant or at least weak. This research, by answering this question will allow us not only to fill an important empirical void in the literature but also to understand the salience of recent class-related discussions in daily Turkish politics and to estimate class-based vote flows in future elections.
LITERATURE REVIEW
Social class struggle was defined as the ‘motor of history’ and has long been instrumental to our understanding of politics. It has been a focus of great interest for its political implications in the Western political science literature. It is suggested that occupational class position, along with class indicators such as education and income, has implications for a wide range of political attitudes and behaviors, including voting preference. A large volume of published studies pays particular attention to the strength of the link between class position of individuals and their voting preference. Much of the previous research investigated the class-vote link quantitatively by exploring its trajectories in the industrialized Western countries. On the other hand, a few theoretical discussions on class voting were largely shaped around class alignment, dealignment, and realignment discussions. The generally accepted assumption of these studies is that working-class members of society, owing to their lower wages, less secure job positions, less autonomy in decisions, less flexible working hours, and more monitoring, are exposed to different working conditions than their middle-class counterparts which in turn effects their voting decisions [2]. Class voting has been an object of research since the 1960’s. Robert R. Alford was the first researcher to propose a standardized measure of class voting. Alford’s analysis translated a greater number of occupational locations into a smaller number of class categories. His decomposition of the social classes was based on the classical Marxist-Weberian social class distinction and produced a two-by-two manual-non-manual/left-right class matrix. In this categorization, while left-wing political parties were representatives of the manual, right-wing ones were of the non-manual classes. His formula was simple: the subtraction of the percentage of non-manual workers who voted for left-wing political parties from the percentage of manual workers who voted for these parties. With this formula, Alford established the traditional axis between non-manual and manual classes and his operationalization of class positions has become standard for class-vote studies for several decades. Alford, himself, analyzed class-voting behavior in four Western democracies—Australia, the United Kingdom, Canada, and the United States—in a dichotomous manual-nonmanual fashion. He found a decline in class voting in these countries between the years 1936 and 1962 [3]. Alford’s collapsing of several occupational categories into two social classes and investigating their voting behavior systematized class-vote analysis and allowed researching class in a comparative fashion. Thus, the topic has attracted considerable scholarly attention. Using Alfordian analytical strategy, many first-wave class-vote analysts attempted to predict class-based voting behavior. While Andeweg carried a class-vote strength analysis research for Dutch voters [4], Listhaug conducted a similar study for Norway [5] and Abramson and his colleagues for the United States [6]. Seymour Martin Lipset presented in his analysis a declining trend for class-vote link in Britain and Germany between 1945 and 1980 [7]. It was followed by many researchers studying class-vote links in Western democratic countries [8], [9], [10], [11], [12], [13], [14], [15], [16], [17], [18], [19]. The main argument in these works is that social class’ importance is gradually decreasing in politics and class-vote link is on decline.
On the other hand, another group of scholars suggested that class voting is not actually on a declining trend [20], [21], [21], [22], [2], [23], [24], [25], [26], [27], [28]. They argued that what makes it look like a declining trend is Alford’s classification of the social class distinction. Alford’s index was sensitive to the popularity of parties, and as left-wing parties gained further electoral support in the post-Cold War era, his index failed to predict the strength of the class-vote link. More recently, Goldthorpe and his colleagues proposed a more nuanced social class schemata [29]; [30]; [20]. They challenged Alford’s ideologically oriented, relatively simple choice of social classes and aimed for further distinction between them. Their social class distinction was based on differences between four social classes: the petty bourgeoisie (self-employed and small employers), the service class-salariat (professional and managerial groups), the routine non-manual (white collar workers), and the working class (skilled, semiskilled, and unskilled manual workers, foremen, and technicians). Using his schemata, Goldthorpe and his colleagues suggested that class remains as an important factor in politics and that class voting is not actually on decline. Many other empirical findings in the field supported this view. Evans and Heath found a similar linear trend in Britain [31], [25]. According to them, what has changed is class size, but class distinctiveness remains important [2]. On the other hand, Alford’s analysis techniques based on the simple formula of, —the subtraction of the percentage of non-manual workers voting for left-wing political parties from the percentage of manual workers voting for left-wing political parties— was also challenged. Researchers argued that instead of measures of absolute class voting, measures of relative class voting such as odds ratios and log-odds ratios should be used. These measures were suggested to be superior to the Alford index and can measure class-vote strength regardless of the changing popularity of the parties and size of social classes. Using measures of relative class voting such as log-odds ratios and Goldthorp (EGP) class schema and data for twenty Western industrialized societies between 1945 and 1990, Nieuwbeerta did not find any general decline in class voting [26]. A decline in Scandinavian countries and, to a lesser extent, in Germany and Britain was evident, but this was far from being an overall one. Evans categorized several potential reasons suggested by the critics of the view that class voting is on the decline in industrialized societies in the West. Author suggested five alternative explanations previously raised for the class-voting decline: 1. Embourgeoisement of the working class 2. Cross-cutting post-industrial cleavages 3. A general increase in education and cognitive skills 4. A value-based political agenda 5. Centripetal movements of the parties [21]. In fact, Evans and his colleagues’ most significant contribution was their supply-side argument. Their argument aimed to reveal whether a top-down or bottom-up mechanism was at work in the class-vote link. Early researchers held the buttom-up sociological mechanism responsible for the decline in class voting. According to them, the de-industrialization of western economies, decline in trade union membership, and rising education and cognitive skills caused a transition from an industrial to a post-industrial society. This paved the way for a series of consequences, including dealignment and a decline in class voting. Criticizing this explanation, Evans and his colleagues proposed their supply-side argument, suggesting that as the working class shrank, left-wing political parties targeted middle-class votes and proposed according policies, which caused centripetal tendencies to emerge in Western politics. Catchall parties declined the tone of class-related signals they sent to voters and in return, voters do not use anymore their social class locations as the main reference for their voting behavior [32]; [21]; [23], [33]; [24]. Empirical findings revealed a mild negative correlation between the convergence of party programs and class voting strength [34].
DATA AND METHODS
This research employed four waves of the World Values Survey (WVS) data to test class-vote link in Türkiye. The WVS data contains all the questions of interest including occupational status, which was translated to social class [20], along with class indicators such as education and income, also religiosity, age and gender as well as question asking about voting choices. Wave dummies were also controlled to capture period effects.
Occupation was used to determine social class position, as it is suggested that while self-reported social class is less reliable than occupational position [2]. Occupational position was translated into three social classes: upper middle class, middle class and working class. The education level variable was used in its original form. It holds eight categories, ranking from the lowest level—inadequately completed elementary school—to the highest one—a university with degree. The income variable was a 10-digit variable holding values ranking from the lowest first step to the highest tenth step. It was recoded into a three-category variable for easier interpretation. Similarly religiosity variable was recoded into three categories indicating low, middle and high level of religiosity. While age was used in its original continuous form, gender is used in its natural binary fashion. Finally the dependent variable is the respondents’ voting preferences. Choices of electorates of three major political parties were included in the data. Owing to the tripartite structure of the dependent variable and no rank order between the voting choices of the respondents, Multinomial Logistic Regression Model (MLRM) is preferred [35].
| Table 1. Descriptive Statistics of Variables | ||
|---|---|---|
| Variable | N. of Respondents | Percent |
| Party Choice | ||
| AK Parti | 478 | 38.33 |
| CHP | 475 | 38.09 |
| MHP | 294 | 23.58 |
| Occupational Social Class | ||
| Working class | 672 | 53.89 |
| Middle | 338 | 27.11 |
| Upper Middle | 237 | 19.01 |
| Education | ||
| Inadequately completed elementary | 33 | 2.65 |
| Completed elementary | 299 | 23.98 |
| Incomplete secondary (technical/vocational) | 60 | 4.81 |
| Complete secondary (technical/vocational) | 106 | 8.50 |
| Incomplete secondary (university-prepatory) | 97 | 7.78 |
| Complete secondary (university-prepatory) | 357 | 28.63 |
| Some university without degree | 23 | 1.84 |
| University with degree | 272 | 21.81 |
| Income | ||
| Low | 382 | 30.63 |
| Middle | 657 | 52.69 |
| High | 208 | 16.68 |
| Religiosity | ||
| Low | 349 | 27.99 |
| Middle | 338 | 27.11 |
| High | 560 | 44.91 |
| Age | 37.16 (mean) | 12.79 (std. dev.) |
| Gender | ||
| Male | 958 | 76.82 |
| Female | 289 | 23.18 |
| Data: World Values Survey, rounds: 3/4/5/6. | ||
| N. of Obs: 1297. | ||
Analyses
The first chart in the figure holds data for the upper middle class. As can be seen from the first part, from the third to the fourth leg, the AK Parti protected its upper middle class vote share around 45 percent. The CHP’s vote decreased from 61 to 44 percent from the first to the second round. In the third round it received 40 and in the fourth 36 percent of upper middle class votes. More fluctuations can be observed in the MHP’s middle class votes. It increased from 38 to 55 from the first to the second round and then decreased to 14 in the third and set on 16 in the fourth round.
The second chart shows middle class vote shares of parties. The AK Parti’s middle class vote decreased from 47 to 40 percent from the third to the fourth round. Similar to the trend in the upper middle class the CHP’s vote share decreased steadily from the first to the fourth round. It was 72 percent in the first round and 46 in the last. MHP on the other hand started off from 27 percent, increased to 47, and dropped to 5 and then 13 percent levels.
In the third chart where working class vote shares are shown, the AK Parti largely protected its working class share around 59 percent across the two last rounds. The CHP’s vote share steadily decreased from 48 to 32 and then to 18 and it increased to 32 in the last round. Lastly, the MHP working class shares are; 52, 67, 21 and 8, in the respective rounds.
Table 2 presents three Multinomial Logistic Regression models predicting preference of each political party over the others. The working-class is selected as the reference category in the occupational class variable. Inadequately completed elementary school is selected as the reference category for the education variable. The income variable is recoded from its original ten-category continuous form into a tripartite one and low-income level was set as the reference category. Similarly religiosity is recoded into three-category format and low religiosity was set as the reference. Age and Age squared as well as gender and dummies for four waves of the data were included as controls.
| Table 2. Multinomial (Polytomous) Logistic Regression Model Predicting Relative Party Preferences by Social Class | ||||||||
|---|---|---|---|---|---|---|---|---|
| AK Parti (relative to the CHP) | MHP (relative to the CHP) | AK Parti (relative to the MHP) | ||||||
| RRR | SE B | RRR | SE B | RRR | SE B | |||
| Occupational Class (Ref: Working) | ||||||||
| Middle | 0.83 | (0.19) | 0.69 | (0.17) | 1.21 | (0.33) | ||
| Upper Middle | 0.60* | (0.13) | 0.69 | (0.16) | 0.86 | (0.22) | ||
| Controls | ||||||||
| Education (Ref: Inad.comp.Ele.) | ||||||||
| Completed elementary | 0.24* | (0.17) | 0.93 | (0.71) | 0.25 | (0.21) | ||
| Incomplete secondary | 0.44 | (0.38) | 1.57 | (1.31) | 0.28 | (0.26) | ||
| Complete secondary | 0.16* | (0.12) | 0.84 | (0.67) | 0.19 | (0.16) | ||
| Incomplete secondary | 0.10** | (0.08) | 0.65 | (0.52) | 0.15* | (0.13) | ||
| Complete secondary | 0.08** | (0.06) | 0.46 | (0.35) | 0.19 | (0.15) | ||
| Some university without degree | 0.02*** | (0.02) | 0.75 | (0.70) | 0.03** | (0.04) | ||
| University with degree | 0.07*** | (0.05) | 0.59 | (0.46) | 0.11* | (0.10) | ||
| Income (Ref. Low) | ||||||||
| Middle | 1.17 | (0.26) | 0.95 | (0.20) | 1.23 | (0.30) | ||
| High | 1.06 | (0.29) | 1.19 | (0.36) | 0.89 | (0.28) | ||
| Religiosity (Ref. Low) | ||||||||
| Middle | 2.97*** | (0.63) | 3.05*** | (0.75) | 0.97 | 0.28 | ||
| High | 12.28*** | (2.90) | 13.28*** | (3.36) | 0.92 | 0.27 | ||
| Age | 0.88** | (0.03) | 0.89** | (0.03) | 0.99 | 0.04 | ||
| Age2 | 1.00* | (0.00) | 1.00 | (0.00) | 1.00 | 0.00 | ||
| Gender (Ref. Male) | ||||||||
| Female | 1.02 | (0.21) | 0.99 | (0.24) | 1.02 | (0.27) | ||
| Wave Dummies | Yes | Yes | Yes | |||||
| Intercept | 66.86*** | (69.78) | 2.57 | (2.80) | 26.00 | (30.74) | ||
| N. of Obs: | 1.247 | |||||||
| Pseudo R2 | 0.3190 | |||||||
| Data: World Values Survey, rounds: 3/4/5/6. | ||||||||
| Significance levels: *p<.05, **p<.01, ***p<.001. | ||||||||
As it can be seen in the first model predicting AK Parti preference over CHP, only upper middle social class is statistically significantly less likely to prefer AK Parti to CHP relative to their working class counterparts (p=0.027). There seems to be no statistically significant difference between the working and middle as well as between the working and upper middle classes regarding to their preference of the MHP over the CHP (p = 0.136 and 0.130, respectively). Similarly there is no statistically significant difference between the likelihoods of preferring AK Parti over the MHP for working and middle as well as working and upper middle classes (p=0.475 and p=0.589, respectively). Education seems to exert important significant differences especially AK Parti preference over the CHP. Educated respondents are on average are less likely to prefer AK Parti over the CHP. We cannot observe any significant differences when we compared inadequately completed elementary school and the other educational categories for their MHP preference to the CHP. Yet, some differences can be observed between the reference category and incomplete secondary, some university without degree and university with degree categories for their preference of the AK Parti over the MHP. The income variable seems to exert no statistically significant effect on the preference of any political party over the others. In other words, higher levels of income in general do not make people to prefer any of the three political parties over the other two. Religiosity seems to have a statistically significant, strong and a linear effect on party preference. Middle level religious respondents who attend to religious services once, on special holidays and once a year are about 3 times more likely to prefer the AK Parti to the CHP than those who attend less often or never (p<0.000). Similarly, high level religious respondents, who attend to religious services once a week or more than once a week are 12.28 times more likely to prefer the AK Parti over the CHP than those who attend less often or never (p<0.000). When we look at the MHP preference over the CHP, the scene is quite the similar. The difference between the middle and low is 3.05 and the difference between the high and low is 13.28 times (p<0.000 for both). Some age differences can be observed as older respondents are more likely to prefer the CHP over the AK Parti and the MHP, yet there is no statistically significant age effect on the AK Parti preference over the MHP. Lastly there is no statistically significant difference between gender categories across preferences of any binary party combinations.
Conclusion and Discussion
Class-vote link has been a widely theorized and tested behavior in the Western political science literature. Analysis of the relationship between social classes and voting preferences has been instrumental to our understanding of voting patterns. The main assumption is that left-wing parties are more likely to appeal to the electoral interests of the lower social classes. One of the recent questions pertaining to the topic is that whether politics is no longer polarized around social class divisions. Findings are mixed and the discussions are still continuing. One thing is evident. Previously published studies are largely limited to research conducted on Western societies and for this reason, much uncertainty still exists about the relationship in non-Western countries. To fill this gap, this research aimed to project class-based party preference in Turkey. It was designed to explore whether Turkish voters’ voting preferences are being affected by their social class locations. To my knowledge, this is the first empirical analysis of the class-vote relationship in Türkiye. Several developments in recent years strengthen the general belief that social class position is gradually becoming an important factor in Turkish politics. Nonetheless, my expectation was that social class do not play a significant role in Turkey and findings have approved this expectation. Having all the other potential variables controlled for, it was revealed that the link between social class location and voting preference is not evident in Türkiye. There may be several reasons for this result. The strong effects of cross cutting divisions such as religious/secular and conservative/progressive might have hampered the class effect. Moreover, ethnicity; kinship and fellow-townsmenship might have also prevented the class structure to be settled [36]. On the other hand, the AK Parti’s and the MHP’s catchall strategies may have worked. It is out of the scope of this research to specify the reason and requires further historical, philosophical and empirical investigations.
Yet, several implications can be achieved from these results. For instance, the AK Parti presents itself as a conservative central political party but it seems that it is more able to appeal working class than the socdem CHP. This finding alone challenges the basic argument of the class-vote research, which suggests that working-class individuals, relative to their middle class counterparts are more likely to vote for left-wing political parties. Based on this result, we can raise the following question: Is this an exceptional situation indigenous to the AK Parti’s catchall strategy? Not, seemingly, as the MHP, a right-wing nationalist party, is not any weaker than the CHP in terms of its working class electoral appeal. In other words, the CHP is not able to attract working-class electoral attention any more than both the AK Parti and the MHP. This may be the underlying reason of the CHP’s new strategy to secure working class votes. Another important implication is that the AK Parti’s electoral success does not only rely on the electoral support of a particular social class but of all classes. There is no social class the AK Parti received less than 30 percent of votes from in any round. With this finding, we can confidently define the AK Parti as a “classwise catchall party”. As a conservative political party, which appeals more working class votes than a deep rooted Social Democratic political party, the AK Parti deserves this qualification. The AK Parti’s populist, cross-class, and inclusive policies seem to have been worked. The head of the AK Parti and President Recep Tayyip Erdoğan’s personal charisma can also be at work here. In a society with low social class intervention in politics, his charisma could have filled the void. Findings particular to the AK Parti could also hint at supporting the supply-side argument. A conservative, center-right political party can easily attract the electoral interest of the working class with the correct catchall top-down signals and the leader’s personal traits. Nevertheless, this idea surely requires further empirical proof.
On the other hand, the MHP’s increasing share of working-class votes between the 1994 and 2004 is remarkable. As a right-wing political party, the MHP seems to be successful in appealing to the working-class electorate before the AK Parti was founded. We can confidently argue that the AK Parti’s entry into the Turkish political scene have reduced the CHP’s and the MHP’s working class electoral support. Perhaps the most surprising finding is the CHP’s failure to attract workers’ electoral attention. Nevertheless, this is obviously a failure for a left-wing social democratic political party and needs to be analyzed carefully to find out the underlying reasons.
This research revealed class-based voting behavior to some extent. Research exploring factors affecting Turkish voters’ electoral decisions has been a growing area of research. It answers some questions, but many still remain unanswered. Class bases of non-voting, strategic and tactical voting, and non-voting, as well as class bases of local and presidential voting, are standing as potentially fruitful research topics in the area. Lastly, it should be noted that the time span of the available data do not reach to 2020’s, which could well be viewed as a limitation.