The difference of bias-corrected daily extreme precipitation in 21st century simulations among GCMs, RCP scenarios, ensemble runs or bias-correction methods are focused in this study. We use 8 GCMs, 2 RCP scenarios and 3 bias-correction methods to estimate the future precipitation. The estimates obtained are compared each other and the degree of the difference caused by the selection of each component to estimate are evaluated. In addition to these three components, the difference of three ensemble runs from only MIROC5 is used to evaluate. The evaluation is conducted in the historical as well as future simulations. The results indicate that the difference is not significant in the historical, but it is significant in the future. We applied a clustering method to estimates in the future. The clustering obtained indicates that the difference caused by the difference of bias-correction methods is comparable to that caused by GCM, RCP or ensemble runs.