Individuals with higher levels of compulsivity exhibit poorer performance on tasks that require model-based planning but the underlying causes have yet to be established. Here, we sought to determine whether these deficits stem from impoverished action-outcome relational knowledge (i.e. issues building an accurate model of the world) and/or an inability to translate models into action. 192 participants performed a two-step reinforcement learning task with concurrent EEG recordings. We found that representations of task-relevant action-outcome associations reflected in reaction time and parietal-occipital alpha-band power were stronger in individuals whose decisions were more model-based, and critically, were weaker in those high in compulsivity. At the time of choice, mid-frontal theta power, a general marker of cognitive control, was also negatively associated with compulsivity, but not model-based planning. These data suggest that model-based planning deficits in compulsive individuals may arise from failures in building an accurate model of the world.