Prior work provides conflicting evidence on whether macro-finance variables can be used to improve predictability of aggregate volatility relative to the naïve benchmark. This paper contributes to this literature by introducing boosted neural networks as a novel statistical approach that learns from its errors and incorporates nonlinearity. This technique is utilized to reexamine the forecasting ability of macro-finance variables for market volatility. The findings show that out of sample predictability is significantly better when the proposed method is used, relative to the alternative approaches used in the literature, including the naïve benchmark, regardless of the state of the economy.

中文翻译：

---

通过增强的神经网络预测总体市场波动性


关于宏观金融变量是否可用于提高相对于朴素基准的总波动率的可预测性，先前的工作提供了相互矛盾的证据。本文通过引入提升神经网络作为一种新颖的统计方法，从其错误中学习并结合非线性，为这些文献做出了贡献。该技术用于重新检查宏观金融变量对市场波动的预测能力。研究结果表明，相对于文献中使用的替代方法（包括朴素基准），无论经济状况如何，使用所提出的方法时，样本外可预测性都明显更好。