结构不稳定影响汇率波动的预测吗?——基于加权极大似然模型平均预测方法

Does Structural Instability Influence Volatility Forecast of the Exchange Rate?——A Weighted Maximum Likelihood Model Averaging Forecast Method

针对汇率波动率潜在的结构不稳定,本文提出加权极大似然模型平均(WMLMA)预测方法,并利用该方法对人民币、英镑、欧元与日元汇率的日波动率进行预测。通过选取特定的衰减参数,WMLMA预测对距离预测期越远的观测值赋予越小的权重以弱化汇率市场不稳定对最终预测的影响,从而使得预测结果能够更为准确地反映预测期数据的潜在结构。实证结果表明,上述四种汇率的波动率均存在显著的结构性变化;相较于基于单一模型、模型选择或模型平均得到的预测结果,WMLMA预测方法取得了更高的预测精度。

This paper proposes the weighted maximum likelihood model averaging(WMLMA)forecast method and applies such method to the volatility forecast of the exchange rates of CNY,GBP,EUR and JPY.Through selecting certain shrinking parameter,the WMLMA forecast assigns smaller weights to the observations that are far from the forecast periods so that their impacts on the final forecast are decreased.WMLMA forecast method makes the final forecast closer to the true structure of the latent data generating process.Empirical results indicate that,the exchange rates under investigation all suffer from structural breaks.Moreover,the forecast based on the WMLMA leads to higher forecast accuracy compared with the forecast based on a single model,model selection,or model averaging.