Release lag of turnover indices shortens due to introduction of time series modelling

The release lag of turnover indices for manufacturing, construction and services shortens by one month starting from data concerning January 2019. In future, all turnover indices are released at a lag of around 45 days from the end of the statistical reference month. The revision also concerns more accurate industries of trade. Time series modelling is used in future to help calculations of the latest month as it enables faster production of high-quality data. In some industries the latest data may become revised significantly in later releases, but these industries are marked separately in the database tables. Time series modelling is also used for some industries at a lag of 75 days in calculations of data released. The retail trade flash estimate, which is released at a lag of just under one month, will remain unchanged.

Depending on the industry, the following methods are used in time series modelling:

1) The ARIMA model, to which factors formed with the main component analysis are combined from the sales inquiry data or the data used by the wage and salary index.

2) Regularisation models (Ridge, Lasso, Elastic-Net), to which factors formed with the main component analysis are combined from the sales inquiry data or the data used by the wage and salary index.

3) The static regression model, which utilises the sales inquiry data or the data used by the wage and salary index.

More information about the used forecast methods is available on the Methodological report.


Last updated 14.2.2019

Referencing instructions:

Official Statistics of Finland (OSF): Turnover of trade [e-publication].
ISSN=1799-0939. Helsinki: Statistics Finland [referred: 23.5.2019].
Access method: http://www.stat.fi/til/klv/klv_2019-02-14_uut_001_en.html