We suggest that, beside the classical endowment and disappointment effects, also envy can be modeled as Kahneman-Tversky loss aversion in a reference-dependent preference framework. In a risk-free setting especially suited to test for envy, a decomposition of referencedependent utility into basic utility and a disappointment evaluation function is achieved, where basic utility is recovered by diagonalization. This feature sets our approach apart from disappointment theories of Sugden (JET 2003) and Bleichrodt (JMP 2009). Empirical evidences on the ultimatum game confirm this model and shows the presence of both inequality aversion on one’s disadvantage (envy), and on one’s advantage (Bellamare et al. Econometrica 2008). We propose a common measure of loss aversion and envy for reference dependent preferences, which generalizes the index of Köbberling and Wakker (JET 2005). The class of constant loss aversion/envy coincides with that of Kahneman and Tversky. We show that Arrow-Pratt absolute risk aversion increases by the relative marginal loss aversion. This inspires further tests of the loss aversion theory of envy.