With recent legislation mandating that publicly traded corporations submit their CEOs' compensation for a non-binding shareholder vote, a systematic understanding of shareholder preferences has never been so important. In spite of this, we know relatively little about what impacts shareholders' preferences and, subsequently, their ultimate voting behavior. We integrate two theories in order to help frame the question and to help predict shareholder behavior. Agency theory predicts that shareholders, as principals, will disapprove of high CEO rewards and poor firm performance, symmetrically assessing gains and losses. Prospect theory predicts that shareholders will be loss averse, responding much more strongly to being in a loss position than to being in a gain or neutral position. We combine these theories' predictions in two lab experiments in which we simulate a shareholder say-on-pay vote, hypothesizing that shareholders will be concerned with agency costs, but only from a loss position. The results of these simulated votes suggest that shareholders do value pay-for-performance, consistent with agency theory. However, shareholders exhibit this focus on agency-normative prescriptions asymmetrically, showing loss aversion consistent with prospect theory. This finding has significant implications for both theory and practice as shareholder votes become a regular and high-profile occurrence.