Low Default Portfolio的calibration 和 回测检验 问题是新巴塞尔模型开发和监控领域的一个非常棘手的问题。其原因为缺乏足够的历史违约数。这篇文章提出了一个保守的算法，规避了历史违约数不足或根本没有的缺陷。

这是英文介绍。

Low default portfolios are those for which banks have little default history, so that average observed default rates might not be reliable estimators of default probabilities (PDs). A key concern for regulators is that credit risk might be underestimated as a result of data scarcity. This paper proposes a quantitative approach to produce conservative PD estimates. Under stylised assumptions about the level of correlation between the risk factors relating to different obligors, and between the systematic risk factor in consecutive years, the estimate of portfolio-wide PD is determined by the size of the portfolio, the number of observed defaults, and the level of confidence that is placed on this empirical evidence. This central PD estimate is then used to adjust the PDs that the bank's own credit experts will have allocated to individual grades within the portfolio. The paper's original purpose was to stimulate debate within the industry on this topic. The revised paper is being issued in the hope that it may assist practitioners in this field.