Nowcasting using linear time series filters

21/03/2018 11:15
This paper focuses on a nowcasting technique which uses a linear decomposition to separate out trend, seasonal influences, and noise (discussion paper of Perrucci and Pijpers, 2017) to
facilitate a forward extrapolation of the trend and seasonal components, including an estimation of the confidence interval. The method is demonstrated using a time series for numbers of unemployment benefits recipients in the Netherland.
In the production of official statistics it is a common problem that there is a time lag between the date on which collection of source data is finalised and the date on which all administrative and technical processes of quality assurance and of data cleaning are completed. This is true both for survey-based statistics and for register based statistics. This time lag can be sufficiently large that a requirement for timely production of statistics cannot be met without some form of forward extrapolation intended to produce official statistics as they are expected to be at the date of publication. This is commonly referred to as 'nowcasting' of time series. This paper focuses on a technique which uses a linear decomposition to separate out trend, seasonal influences, and noise to facilitate a forward extrapolation of the trend and seasonal components, including an estimation of the confidence interval.
The method is demonstrated using a time series for numbers of unemployment benefits recipients in the Netherlands, available on the Statistics Netherlands website.

Downloads