This paper uses the conditional VaR to derive the optimal portfolio investment weight of the bond, which is the main investment target of special account of the mutual finance of NationalAgricultural Cooperative Federation(NACF). Also, this paper compares the optimum portfolio weights of the mean-variance model and the conditional VaR model. Based on the derived results, we confirm whether the predicted return distribution is close to the normal distribution or not. In this study, the Bayesian approach is applied to estimate the posterior predictive distribution. The Bayesian method can easily reflect uncertainties in parameters and models. A vector-autoregressive model was used to estimate the posterior predictive distribution using the Bayesian method. For an empirical application, we use data on daily returns for government bond, industrialfinancial debentures, corporate bonds. When the Bayesian method is used to estimate the posterior predictive distribution, the results of the two models are derived similarly. Also, we find that the optimal investment weights vary depending on the target return rates.