A statistical analysis of creep rupture data for 2.25Cr-1Mo steel was performed. The scattering of creep rupture data was represented by Z-parameter method based on Manson-Haferd method. With the application of Z-parameter, reliability design for allowable stress of creep rupture strength was carried out according to design life. The higher the value of confidence level, the lower the allowable stress. In comparison with safety factor method and minimum rupture strength method, it can be seen that reliability design based on Z-parameter is more agree with experimental data than other methods. Reliability design provides more precise results by considering the real distribution of creep rupture property and provides more flexible choice for design due to the need of safety and economy.