Beliefs, economics, and spatial regimes in voting behavior: the Turkish case, 2007–2018

Turkey's political landscape has been significantly shaped by the AKP's rise to power in 2002. Despite economic setbacks in subsequent terms, the AKP consistently won general elections, prompting questions about the drivers of its sustained public support. While the Economic Voting Theorem (EVT), encompassing both pocketbook and sociotropic voting, is often cited, it fails to fully explain the AKP's dominance. This study addresses this gap by incorporating spatial components and peripheral sociologies (Center-Periphery theory) to analyze voting patterns. The research aims to empirically determine the relative contributions of EVT and C-P factors in shaping AKP and pro-Kurdish party vote shares across different regions of Turkey. Understanding this will shed light on the broader dynamics of electoral behavior in illiberal regimes facing economic challenges.

The literature on voting behavior commonly explores the link between economic conditions and voting patterns. Egotropic (pocketbook) voting emphasizes voters' assessments of their personal economic circumstances, while sociotropic voting focuses on the national economic climate. Both are aspects of EVT. Studies show varying levels of influence for these factors. However, alternative theories highlight the role of structural factors like religion and ethnicity. The Center-Periphery (C-P) framework helps explain voting behaviors among peripheral groups based on their identities and geographic locations. Lipset and Rokkan's work emphasizes territorial and cultural conflicts between central and peripheral forces, shaping party systems and voter alignments. The Turkish context exhibits such cleavages, particularly the tension between secularist central forces and peripheral groups defined by religious conservatism and Kurdish identity. Previous research on Turkish elections has used EVT or C-P, but few have integrated both with a spatial perspective.

This study uses a comparative method incorporating EVT and C-P theories. The analysis employs an original dataset spanning 81 Turkish provinces from 2007 to 2018, covering five general elections. The dataset includes: * **EVT variables:** National and local GDP growth rates, unemployment rates, and inflation rates (weighted for four quarters before each election). These capture both pocketbook and sociotropic voting effects. * **C-P variables:** A proxy for religious conservatism (attendance in Qur'an courses), and a spatially interpolated estimate of the Kurdish population percentage at the provincial level. These capture ethnic and religious identities. * **Control variables:** Median age, higher education graduation rates, party depreciation (lagged vote share and years in power), and strategic voting (lagged vote share). These account for socioeconomic and political factors. * **Outcome variables:** Vote shares of the AKP and pro-Kurdish parties. A regression model is used to assess the impact of EVT and C-P variables on AKP and pro-Kurdish party vote shares. Spatial analysis, specifically Local Indicators of Spatial Association (LISA), identifies clusters and outliers to reveal spatial regimes in voting behavior. A spatial regime model with a Chow test analyzes differences in coefficients across regimes defined by Kurdish population percentages (above or below 15%). This allows assessing whether EVT and C-P effects vary across these spatially distinct voter groups.

The regression analysis reveals that for the AKP, sociotropic voting (national economic conditions) is a stronger predictor of vote share than pocketbook voting (local economic conditions). Contrary to expectations, local economic variables (growth, unemployment, inflation) are insignificant predictors of AKP vote share nationwide. However, C-P factors, particularly religious conservatism (Qur'an course attendance), significantly and positively influence AKP vote share, even exceeding the impact of national economic growth. Kurdish population size negatively impacts AKP vote share, but less strongly than religion. Demographic factors (median age, higher education rate) are also significant predictors. Spatial analysis reveals distinct spatial regimes based on Kurdish population density. A Chow test shows significant structural breaks across regimes (Kurdish population above/below 15%). In Regime 0 (lower Kurdish population), pocketbook voting (via unemployment) is observed alongside sociotropic voting. In Regime 1 (higher Kurdish population), only sociotropic voting is significant. In both regimes, C-P factors, notably religion, show greater influence than EVT. Pro-Kurdish party vote share analysis shows that in Regime 0, national economic growth negatively affects their vote share, suggesting that AKP's economic successes may draw voters from Pro-Kurdish parties. Kurdish population size positively and significantly impacts Pro-Kurdish vote share in both regimes, indicating that ethnic identity is a primary driver, even overriding economic concerns. Religion negatively impacts Pro-Kurdish vote share in both regimes. Spatial diagnosis suggests distinct spatial processes within each regime, implying no spillover effects across regimes.

The findings challenge the conventional emphasis on EVT in explaining the AKP's electoral success. While sociotropic voting exists, the stronger influence of C-P factors, especially religious conservatism, highlights the importance of non-economic factors in shaping voting behavior. The existence of distinct spatial regimes underscores the heterogeneity of voting patterns across Turkey, with ethnic identity playing a crucial role. The results suggest that the AKP's success is grounded in a combination of economic performance at the national level and the mobilization of religious conservatives at the local level. This mobilization appears particularly successful in areas with lower Kurdish population density. The lack of significant local economic effects for the AKP may reflect targeted government spending or other strategies which counterbalance local economic hardship. Future research could explore these strategies in more detail. The contrasting responses of the AKP and Pro-Kurdish parties to economic and identity-based factors highlight the complexity of political dynamics and cleavages within the Turkish context.

This study reveals that the AKP's consistent electoral success is rooted in a complex interplay of national economic performance (sociotropic voting) and the mobilization of religiously conservative voters, particularly in areas with lower Kurdish populations. Ethnic identity is a powerful driver for Pro-Kurdish party support. The study highlights the importance of incorporating both EVT and C-P frameworks, along with a spatial dimension, to fully understand voting behavior in diverse contexts like Turkey. Further research could investigate the specific government policies that influence voter perception and the dynamics of spatial spillover effects across different identity groups.

The study relies on proxies for certain variables (religious conservatism, Kurdish population) due to data limitations. The spatial interpolation of Kurdish population data introduces uncertainty, although efforts were taken to mitigate this. The model could be further refined by incorporating additional variables that capture deeper aspects of voter preferences and political dynamics. While the spatial analysis identifies distinct regimes, it doesn’t fully explore the complex interactions within those regimes.