A monthly estimation method of ILO unemployment: a state-space framework

The International Labour Organization (ILO) defines unemployed population as the persons who are not in employment, immediately available to work and who actively look for a job. At INSEE, the mean of the ILO unemployment rate is computed at a quarterly frequency using the Labour Force Survey (LFS). Survey periods are uniformly distributed all over the year; hence a monthly unemployment series can be computed. However, this series is highly volatile because the LFS sample rotation has been designed for computing quarterly data and not monthly data. In the first part, this article describes a state-space filtering method in order to extract from these volatile series underlying trends that are useful for economic analysis. The filter exploits the LFS overlap in order to optimally identify (in a mean squares sense) the sampling noise in the survey. Of course, there remains statistical uncertainty even if this identification is optimal. This method could be extended in order to estimate the unemployment rate on smaller units (sub-populations or geographical areas). In the second part of the article, the statistical model is adapted in order to produce a monthly ILO unemployment rate in the same timing as Eurostat. In this case, the unemployment rate is defined as the rolling quarterly mean of LFS results and the state-space model is used as a forecasting tool.