We study the endowment effect using four incentivized representative surveys with 4,000 U.S. adults. We replicate the standard finding of an endowment effect for lotteries—a divergence between Willingness to Accept (WTA) and Willingness to Pay (WTP)—but document three new findings. First, we find little evidence that the endowment effect is related to loss aversion for risky prospects, contradicting predictions of leading explanations such as prospect theory. Second, WTA and WTP not only diverge but are, at best, weakly correlated. Third, WTA and WTP strongly relate to other aspects of risk preferences. Our results allow us to differentiate between existing theories, and point to Salience or Cautious Utility as the most complete explanations for our data.