Much recent political psychology scholarship has examined the role of anxiety in vote choice. This work generally argues that anxiety affects vote choice indirectly by causing citizens to more thoroughly search for and process political information. This indirect effect of anxiety leads citizens to rely less on heuristics, such as party, and more on substantive information, such as policy positions. The most prominent example of this scholarship is the Affective Intelligence (AI) theory of emotions. In this paper, we use cross-sectional and panel survey data to test AI against two simpler alternatives: (1) that emotions directly influence candidate evaluations and (2) that candidate evaluations directly influence emotions. We first show that these simpler alternatives can produce the complex, cross-sectional interactions that provide the principal support for AI. Then, using panel data to better assess causal direction, we find little support for AI, some evidence that emotions directly influence candidate evaluations, and strong evidence that candidate evaluations directly influence emotions. Scholars, we conclude, should be hesitant to abandon these simpler explanations in favor of AI.