Winning Elections with Unpopular Policies: Valence Advantage and Single-Party Dominance in Japan

Do voters support dominant parties in democracies because of policy preferences or non-policy (valence) factors? We consider the pre-eminent case of Japan's Liberal Democratic Party (LDP), and investigate whether policy preferences or valence can better explain voting behavior in three recent elections (2017, 2021, 2024). We first introduce a new measurement strategy to infer individuals' utility for parties' policy platforms from conjoint experiments. Using this measure, we find that policy preferences positively correlate with vote intentions, but are not sufficient to explain LDP dominance. Many LDP voters in each election actually preferred the opposition's policies. Moreover, the LDP lost support in 2024 despite proposing a more popular set of policies. To understand what accounts for this disconnect, we experimentally manipulate party label and decompose its effect, revealing that trust appears to be an important non-policy variable behind LDP support. We interpret these findings as evidence that much of the LDP's support should be attributed to its valence advantage over the opposition, rather than voters' preferences for its policies.