Bootstrapping structural time series models

This paper presents a method to compute standard errors of estimates based on structural time series models.
Statistics Netherlands applies a structural time series model to compute monthly figures about the labour force. Standard errors of these estimates can be computed by well-known analytic formulas. The monthly figures are also used to derive quarterly and yearly figures and differences between two time periods. For most of these derived figures no analytic formula for standard errors is available. Therefore, a bootstrap algorithm is tested to derive standard errors for these figures.

Downloads