Affective polarization, or dislike and distrust between partisans, is thought to stimulate engagement but harm political trust and commitment to democratic norms. However, recent experimental work has cast doubt on these findings. In this paper, we use longitudinal models test competing possibilities as to why manipulating affective polarization lacks effects on democratic outcomes: that the effect exists but experiments poorly capture it, reverse causation, and spurious effects. Using over a dozen separate panel studies, we find strong evidence that most effects of affective polarization on positive or negative democratic outcomes are spurious. Observed associations using traditional cross-lagged panel models do not remain when using models that perform better at separating between- and within-person effects. This suggests that affective polarization co-occurs with, but does not cause, its purported outcomes.