Good risk measures, bad statistical assumptions, ugly risk forecasts
提出时间异质学生t自回归模型作为波动率预测的替代方案，实证表明该模型预测更优，且风险度量问题主要源于模型风险而非风险度量本身，对监管者和从业者有建议。
Abstract This paper proposes the time‐heterogeneous Student's t autoregressive model as an alternative to the various volatility forecast models documented in the literature. The empirical results indicate that: (i) the proposed model has better forecasting performance than other commonly used models, and (ii) the problem of reliable risk measurement arises primarily from the model risk associated with risk forecast models rather than the particular risk measure for computing risk. Based on the results, the paper makes recommendations to regulators and practitioners.