Since the early 1990s, a 25%-45% gender pay gap has persisted for the top five executives in U.S. publicly traded companies. I present an empirical approach to determine the relative importance of two possible explanations for the gender pay gap that originate with employers: gender taste-based discrimination, and downward-biased beliefs about women's performance. I use the 2003 SEC regulation event that required boards to become more independent and disallowed insiders to serve on the compensation committee to distinguish between the two possible causes of discrimination. Independent board members do not work alongside executives and so would be less inclined than insiders to indulge in taste-based discrimination. Independent board members, on the other hand, have less information about executives' performance and are thus more likely to rely on their prior, potentially biased beliefs about women's performance when they set pay. I find that the gender pay gap became 19% larger in firms that were required to convert to more independent boards compared to firms that were not, which is not consistent with taste-based discrimination. An increase in the pay gap is consistent with downward-biased beliefs about women's performance, but also with reverse taste-based discrimination. I distinguish between these two hypotheses by examining whether the increase in the pay gap is persistent and uniform across positions. I find the increase in the pay gap reverted as independent board members had time to learn about individual performance. And, the gap did not widen in occupations where accreditation provided an easy-to-interpret signal of ability. These results are consistent with board members having downward-biased beliefs about women's performance.