Regression Discontinuity diagnostics reveal statistical anomalies in Turkish elections

Using Regression Discontinuity diagnostics we document a number of statistical anomalies in the 2004 Turkish mayoral elections. The governing party that controls the parliament is much more likely to win close races than lose. Moreover, compared to close governing party losses, there is a sharp drop in turnout and contending party votes in close governing party wins. Finally, the parties that disproportionately lose very close races are exclusively ideological competitors of the governing party. Among the potential mechanisms that may create those anomalies, electoral manipulation seems to a plausible explanation. Those anomalies show that the outcomes of very close popular elections can be non-random and that the assumption of the continuity of the expected potential outcomes at the threshold could be violated. We discuss implications of our findings for Regression Discontinuity Designs and for understanding the consolidation of the right-wing electorate in Turkey during the last decade.