There is a serious inequality of dietary health between the high-income and low-income classes in China, which is generally considered as an important issue facing this country right now. When inferring potential causes of this inequality, extant literature usually neglect the effect of an individual’s taste itself. Leveraging on the panel structure of CFPS 2010-2014 data, we employ recentered influence function (RIF) method to first eliminate individuals’ "inborn taste" through controlling for individual fixed effect, based on which we then conduct unconditional quantile decomposition to further separate the effects that is due to (1) individuals’ different characteristics (e.g. socioeconomic and demographic factors) and (2) individuals’ different marginal utility (i.e. coefficients) of such characteristic even if they hypothetically have the same characteristics, which we define as the effect of "alterable taste". We show that with the presence of income class mobility, "alterable taste" and socioeconomic status can explain 54.3% and 44.5% of the dietary health inequality, respectively; In the absence of income class mobility, the two different tastes can explain 37.1% and 62.3%. These results imply that government’s income policy can only partially influence individuals’ dietary health. The policy makers are suggested to invest more in taste education and helping individuals to foster a heathy and balanced diet.