This article reports and provides an explanation for a discrepancy between two theoretically equivalent, frequently used, and incentive-compatible methods of measuring premia for improved novel products: the full-bidding and endow-and-upgrade methods. We found the following reverse endowment effect in a willingness-to-pay (WTP) elicitation Becker-DeGroot-Marschak (BDM) experiment for the newly developed biofortified high iron pearl millet (HIPM) conducted in rural India. The WTP for exchanging local pearl millet (LPM) for HIPM (the endow-and-upgrade measure of premium for HIPM over LPM), was significantly greater than the difference between the WTPs for HIPM and LPM (the theoretically equivalent full-bidding measure). Our explanation is that subjects who possess an existing version of a product experience a reversal of loss aversion with respect to the novel and improved version of the product. We identify and structurally estimate the reverse loss aversion parameter. Our findings caution against using endow-and-upgrade and full-bidding methods interchangeably for measuring premia for new products, even if the experimental design accounts for reciprocity and experimental-income effect.