Accounting estimates are developed in a bottom-up fashion; subordinates generate estimates that are reviewed by managers. The anchoring heuristic suggests managers may be highly influenced by subordinates’ initial estimates. However, motivated reasoning theory predicts that reporting incentives will bias managers’ review in favor of estimates that are incentive consistent, and managers will selectively attend to information that supports their preferred conclusion, including their perceptions of the subordinate. Using experimental methods I manipulate the consistency of the subordinate estimate with management reporting incentives, and the narcissistic description of the subordinate. Consistent with motivated reasoning theory, I find that managers anchor on incentive consistent subordinate estimates, regardless of subordinate narcissism, but anchor less on incentive inconsistent subordinate estimates, especially when the estimate comes from a narcissistic subordinate. I also find evidence that managers believe narcissistic subordinates act strategically in their own self-interest, and selectively attend to this belief to adjust away from incentive inconsistent subordinate estimates, but not incentive consistent subordinate estimate. My results reveal two potential weaknesses in the management review process: susceptibility to subordinate anchors, and bias created by reporting incentives.