Is it possible to increase one's influence simply by pretending to be confident that one is correct? There are two competing hypotheses in the literature – the confidence heuristic hypothesis (more confidence is always better) and the calibration hypothesis (confidence will backfire if you make even a peripheral mistake). This paper reconciles this discrepancy and argues that people care more about calibration than confidence, but that they will revert to a confidence heuristic when calibration becomes costly. The results of Study 1 show that highly accurate advisors benefit from displaying confidence, whereas low accuracy advisors actually prove more credible when they express less confidence; highly confident but inaccurate advisors received the lowest ratings of credibility. But Study 2 shows that when feedback is unavailable or costly, more confident advisors hold sway regardless of accuracy, and people will seek feedback less often when advisors display high confidence rather than low confidence.