The delay and decline in union formation are of particular importance for understanding low fertility in societies characterized by a strong link between marriage and childbearing. While the so-called marriage market mismatch provides one compelling explanation for later and less marriage, there is a lack of consensus on how to measure partner preferences; moreover, studies that have relied on observational data suffer from a range of potential biases. To circumvent these limitations, we employ a conjoint survey experiment approach to examine how gender asymmetry in partner preferences may play a role in spouse selection by focusing on socioeconomic status in Japan. While our results confirm gender-specific partner preferences, we do not find strong evidence supporting heterogeneous preferences, which is a core assumption of the marriage market mismatch hypothesis. We also find that educational attainment plays a relatively limited role in partner preferences compared with other socioeconomic status measures.