In this paper, we estimate a rich model of college major choice using a panel of experimentally-derived data. Our estimation strategy combines two types of data: data on self-reported beliefs about future earnings from potential human capital decisions and survey-based measures of risk and time...