This paper nds the optimal look-back period for adequate and less procyclical credit capital requirement forecast for the aggregated credit portfolio. Our empirical experiments firstly show that credit capital forecast based on shorter look-back period is adequately lower compare to longer look-back period due to stronger homogeneity of historical scenarios during the non-GFC period. On the contrary, shorter look-back period provides larger credit capital forecast as the GFC occurs, since it is more risk sensitive to newly observed credit events due to its smaller sample size for calibration. Thus, capital requirement forecasts based on shorter-term samples are more adequate, but also become more procyclical compare to longer-term samples. Secondly, capital requirement forecasts based on stand-alone method are more adequate than those reecting diversication due to strong tail dependence especially during the GFC period. Finally, stand-alone VaR and capital requirement forecasts based on 11-years length of look-back period, the average duration of two business cycles in U.S economy, are optimal on the basis of micro-prudential adequacy and macro-prudential less procyclicality for the aggregated portfolio in U.S. banking system.​

Keywords: Unconditional loss distribution, VaR, Procyclicality, Diversication, Copula, Capital requirement.