This paper adopts individual portfolio choice data to estimate the preference parameters in almost stochastic dominance (ASD). ASD is a decision criterion for most decision makers to rank distributions on the basis of riskiness. Existing empirical studies have shown that ASD is helpful in explaining some puzzles in nance and could be applied to evaluate investment strategies. Their conclusions heavily rely on the estimation of the preference parameters in the ASD rules provided by Levy et al. (2010). However, the estimation of Levy et al. (2010) is obtained from arti cial tasks designed in a lab- oratory by using students as subjects and adopts an incorrect condition for the almost second-degree stochastic dominance. Our paper is the rst to use empirical observations in portfolio choice decisions and to adopt the correct conditions to estimate the pref- erence parameters in ASD. Our estimation shows that these preference parameters are much lower than the numbers in Levy et al. (2010). Since these parameters serve as upper bounds to obtain the unambiguous prediction on the ranking of distributions, our ndings suggest the needs of re-examination on the existing empirical conclusions.
American Risk and Insurance Association2018 Annual Meeting