Essentials of Time Series for Financial Applications

when ${\hat{\sigma }}_{t+1|t}^2<{\overline{\sigma }}^2$, that is, starting from a relatively low level of predicted variance (0.0095), as it was the case in US bond markets at the end of 2016. Because in this example, the estimated persistence index $\hat{\alpha }+\hat{\beta }=0.99$ is relatively high, the convergence of predicted variance to ${\overline{\sigma }}^2=0.0144$ takes many weeks, approximately 250 (e.g., $1-{(\mathrm{0.06\; ...}}^{}$