Estimation of level and change for unemployment using structural time series models
Section 2. The Dutch Labour Force Survey

• Previous
• Next

The Dutch LFS is a household survey conducted according to a rotating panel design in which the respondents are interviewed five times at quarterly intervals. Each month a stratified two-stage sample of addresses is selected. All households residing on an address are included in the sample. In this study 72 months of LFS data from 2003 to 2008 are used. During this period the sample design was self-weighted. The first wave of the panel consists of data collected by means of computer assisted personal interviewing (CAPI), whereas the four follow-up waves contain data collected by means of computer assisted telephone interviewing (CATI).

The Netherlands is divided into twelve provinces which serve as the domains for which monthly unemployment figures are to be estimated. Monthly national sample sizes vary between 5 and 7 thousand persons in the first wave and between 3 and 5 thousand in the fifth wave. Provincial sample sizes are diverse, ranging from 31 to 1,949 persons for single wave monthly samples.

LFS data are available at the level of units, i.e., persons. A wealth of auxiliary data from several registrations is also available at the unit level. Among these auxiliary variables is registered unemployment, a strong predictor for the unemployment variable of interest. These predictors are used to compute initial estimates, which are input to the time series models.

The target variable considered in this study is the fraction of unemployed in a domain, and is defined as ${\overline{Y}}_{it}={\displaystyle {\sum }_{j\in i}}{y}_{ijt}/N_{it},$ with $y_{ijt}$ equal to one if person $j$ from province $i$ in period $t$ is unemployed and zero otherwise and $N_{it}$ the population size in province $i$ and period $t.$

• Previous
• Article main page
• Next