Seminarium POLPAN: A new approach to the inconsistency of retrospective data on work and occupation in panel surveys

25 lutego odbyło się Seminarium POLPAN, podczas którego prof. Zbigniew Sawiński (IFiS PAN) wygłosił referat zatytułowany: „A new approach to the inconsistency of retrospective data on work and occupation in panel surveys”.


Longitudinal surveys usually avoid repeating questions about the same facts from the respondent’s biography. This principle also applies to the reconstruction of occupation careers. The panel founder questionnaire usually contains questions covering the period from the first job in life to the current job, while in the subsequent waves of the panel questions about work are limited to the period between the previous and current interviewer’s visit. This approach can be called a „single-measurement” because information about each job is collected only once.

I will present arguments for another approach called „multi-measurement”, which takes into account information about the same jobs, collected in different panel waves. Such information means that some work may have started before the date of the previous interview, therefore the respondent has already been asked about this work in at least one of the previous panel waves. The Polish Panel Survey POLPAN well illustrates the fact that in panel surveys this type of multiple measurement is quite common. POLPAN began in 1988 and was carried out every five years since. In the seventh wave in 2018, almost half of the working panelists stated that their current work was started before the interviewer’s previous visit in 2013. Older panelists participating in the POLPAN survey from the beginning, most of whom were retired, were asked about their last job instead of the current one. Over half of them stated that their last job began before 1988. This means that information about these jobs was collected in each of the seven POLPAN waves.

Using POLPAN data, I will try to answer the basic question: whether the strategy of using multiple measurements does not lead to a significantly different picture of occupational trajectories as compared to a single measurement.

I will also mention some significant benefits that the multi-measurement approach provides. First, it helps to identify occupations that are particularly prone to errors, and probably require more sophisticated interviewing methods. Secondly, the multi-measurement approach allows for taking into account more than one occupation at the same time, which is a useful solution for respondents performing several simultaneous jobs. And the third advantage: work sometimes combines different occupational roles, like a teacher who is also a deputy headmaster. It is difficult to reflect the essence of such work with a single occupational code. The multi-measurement approach seems more promising in this context. All listed benefits will be illustrated with examples from POLPAN.

The multiple measurement approach requires special tools for data analysis. Three such solutions proposed in POLPAN will be briefly discussed. In addition, completely new proposals will be considered. One of them is fuzzy classifications, which have been successfully implemented in many fields of science.

Key words: panel surveys, occupational trajectories, multiple measurement, longitudinal data analysis, the Polish Panel Survey POLPAN