Re-examining the contamination effect of Japan's mixed electoral system using the treatment-effects model

Empirical studies on mixed electoral systems have shown that parties obtain additional PR votes by nominating many SMD candidates (contamination effect). I argue that the existing studies suffer from a methodological problem that stems from the nature of candidate nomination. That is, parties tend to nominate SMD candidates in areas where they are strong, which makes the impact of running candidates look larger than it actually is. This study employs the treatment-effects model that corrects the endogeneity bias and analyzes the results of recent general elections in Japan. Contrary to the accepted view, no statistical evidence is found that supports the existence of the contamination effect, implying that Japan's party system may converge to a two-party competition.