Technical note: Kalman Filter Estimation of the KNW Model

Activity: Consultancy / Advice / PRConsultancy / Advise external partiesAcademic


This technical note gives implementation notes for estimating the Koijen-Nijman-Werker model from historical data based on a Kalman filter. We provide an independent derivation of the KNW model. We propose a different implementation of the state-space formulation of the KNWmodel and we test the impact of two different specifications for the initialisation of the Kalman filter maximum-likelihood estimation. By doing so, we provide an independent verification of the parameter estimations provided by DNB for the Committee Parameters. We find no significant differences between the parameters estimates reported by DNB and our own parameter estimates.
Period28 May 2019
Work forMinisterie van SZW
Degree of RecognitionNational