In 2017, Statbel (Statistics Belgium) introduced a major reform for the Labour Force Survey (LFS): after 18 years of working with a continuous survey, the switch was made to a panel survey. The most important aspects of this reform are: • The transfer to an infra-annual (“quarterly” for the Belgian case) rotating panel design. A sample (called rotation group (RG)) of private households is drawn each quarter, independently of previously drawn samples. The sample is rotating in the sense that any specific rotation group stays in the survey during 18 months (6 quarters), after which it is replaced by a new rotation group. In the panel survey, each member (of at least 15 years old) in a selected household is asked to complete a questionnaire four times, i.e. during four waves, according to a 2(2)2 scenario: a selected household/individual is asked to complete a questionnaire during two consecutive quarters (wave 1 and wave 2), is then not in the survey during the next two quarters, and is again asked to complete a questionnaire during the next two quarters (wave 3 and wave 4). • The introduction of mixed mode data collection techniques. In the first wave and after an introductory letter, the selected households are contacted by an interviewer and CAPI is used to collect the data. In the three follow-up waves, data can be delivered through CAWI or CATI, according to the household’s preference. • Application of the wave approach. Information on structural variables is gathered in the first wave only; information on core variables is collected in all four waves. • A revision of the weighting methods. More attention is paid to the correction of effects of non-response (in the first wave) and panel attrition (in the follow-up waves). This resulted in a two-step weighting approach: in step 1, response probabilities are estimated through a mixed effects logistic regression model, and aggregates of the estimated probabilities are used to correct the sampling weights; in step 2, the corrected weights from step 1 are calibrated to the population of interest. The present text is focussing on the latter aspect of the reform, i.e. the weighting methods. We show the effect of changing the weighting method by comparing estimates for various LFS indicators based on the new 2-step approach, with estimates based on the old 1-step approach which was used for the continuous LFS from 1999 to 2016. We argue that the new method is better correcting for non-response and attrition bias. Furthermore, it will be shown that different changes in the new methodology are, to some extent, cancelling out, causing only moderate breaks in time series for some major indicators.