Empirical evidence show that credit spread changes are not normally distributed, but show properties of heavy-tails, skewness, time-varying volatility, and the possibility of co-movements. These properties if ignored compromises the value-at-risk models in financial risk management. Therefore, the aim is to critically compare and evaluate the performance of the component valueat-risk methodologies for credit spreads in a fixed income portfolio, under different distributional assumptions including the Multivariate normal distribution and the Multivariate student t distribution. The underlying portfolio is a Euro-denominated BBB corporate bond portfolio and therefore the credit spreads are the European BBB credit spreads. The methodologies used are the normal linear value-at-risk and the Monte Carlo value-at-risk where the credit spreads are modelled under a Multivariate normal distribution and the Multivariate student t distribution. The findings present that the European BBB credit spread changes show properties of heavy-tails, skewness, and co-movements. Therefore, the credit spreads modelled under the assumption of normality are concluded to show underestimated value-at-risk estimates of the credit spreads in the portfolio at very low significance levels. However, when with the assumption of being Multivariate student t distributed, the value-at-risk estimates are higher at very low significance levels. Therefore, the appropriate assumption of distribution is dependent on the significance levels. In the evaluation of the methodologies, it is noted that the normal linear value-at-risk method is computationally easy and fast to apply, but the assumption of normality in the distribution can cause inaccuracy in the value-at-risk estimation. Furthermore, the high correlation and the time-varying properties of the credit spreads should be modelled with a heteroscedastic model. The Monte Carlo method is the most efficient to estimate the value-at-risk because of its flexibility in the assumptions concerning linearity, conditional normality, nonnormality, and time variation among others. However, the computational time for institutional portfolios are very large, and the method is expensive in the requirement of investment in infrastructure and intellectual capital.