I have estimated the term premium for the Polish Government Bonds (POLGBs) using the methods described by Adrian et al. (2013) (often referred to as ACM). The underlying yields are interpolated from the Nelson-Siegel-Svensson model based on a bootstrapped zero-coupon yield curve.

I have estimated the model for two time series:

1. Daily 10Y POLGBs Yields from 24/01/2005 to 20/05/2022

enter image description here 2. Monthly 10Y POLGBs Yields from the last day of each calendar month from 24/01/2005 to 20/05/2022 enter image description here

Although the monthly yields are subsetted from the daily yields, the picture painted by each term-premium estimate is drastically different, especially during the 2020-2021 quantitative easing program from the Polish Central Bank. The monthly term-premium would have similar dynamics to the analogous term-premium estimates for the German BUNDs. Which estimate would you find more reliable?

  • 1
    $\begingroup$ At a glance the monthly version is far more sensible. The preferred approach is to estimate the model using monthly data, then apply the parameters to daily data. $\endgroup$
    – Helin
    Commented Jun 28, 2022 at 21:19
  • $\begingroup$ Thanks for the help @Helin. Just to make sure I have got the procedure right - I should estimate the A and B' for the selected maturities on monthly data, then re-run the PCA on the daily yields and construct fitted yields using the factors of daily yields? $\endgroup$ Commented Jun 28, 2022 at 22:27
  • $\begingroup$ @Helin I would assume that a model estimated this way for daily data would have a fit drastically worse than the monthly model. Am I correct? $\endgroup$ Commented Jun 29, 2022 at 19:02


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.