Marginal component effects (MCEs) are the most popular causal estimands in conjoint experiments, but they have been recently criticized on both conceptual and methodological grounds. This debate exclusively considers average MCEs (i.e. AMCEs), but many of the questions raised can only be addressed with individual-level data. In this paper, we contribute to the literature by estimating individual MCEs and exploring their predictive validity. Our evidence comes from two studies, one in Germany and one in the United States, in which respondents both rate hypothetical candidates in conjoint experiments and report their support for real-world parties. We show that respondents’ issue preferences measured via individual MCEs in conjoint experiments strongly and reliably predict real-world party support in the expected directions. This is true independently of various aspects of the conjoint design, such as the number of rated profiles and the nature of the outcome (choice vs. rating). Our findings demonstrate that conjoint MCEs constitute valid measures of preferences.