Based on theoretical analysis of crop risk and premium rate setting, we take the case of premium rate setting of insurance on cotton yield per unit area in Shache County, Shaya County and Aksu City of Xinjiang. Using parametric methods and insurance actuarial technique, we select the optimal model for risk fitting of cotton yield per unit area in three areas; compare the premium rate calculated accurately under four risk distribution assumptions of cotton yield per unit area, and the rational premium rate, to analyze the impact of risk distribution of crop yield per unit area on premium rate setting. The empirical results show that the Logistic distribution is the optimal distribution for risk fitting of cotton yield per unit area in three areas; the rational net premium rate of cotton insurance in three areas is 7.62%, 6.32% and 4.96%, respectively; there are errors in premium rate setting under assumptions of normal distribution, normalized skew distribution and Weibull distribution, ranging from 0.2 percentage points to 8 percentage points. Thus, it indicates that the selection of risk distribution model of yield per unit area directly affects the accuracy of premium rate setting of crops, and the key to accurate premium rate setting of crops lies in correct selection of risk distribution model of yield .