children less than 16) and (iv) the link file that allows to follow in time the longitudinal
status of a given observation unit surviving along the panel time window.
The observation unit considered throughout the paper is the individual, either in cross-
sectional or longitudinal analysis. However, the household remain the unit of measure for
some variables, such as the variable of resources (income), or to characterize the
environment in which the individual lives.
Starting from the original data, some methodological issues have to be dealt with and
choices have to be made concerning the definitions of variables as they have direct impact
on the analysis carried out.
The first issue to be addressed, as it is probably the most important to this paper, is the
definition of retired individual. There are different forms of defining the “retired” state
from micro data, each one with advantages and drawbacks. The various definitions depend
on how the reality is inquired and on which criteria is adopted to configure the concept, of
objective or subjective nature. For instance, considering an objective criteria, we can
classify an individual as being retired if he/she has a pension paid by the social security
system as the main income source or - it is another example - to agree on an exogenous or
conventional classification based on self declared number of hours worked. In subjective
terms, the individual can be classified as retired from each individual’s own assessment of
his or her labour market status. We assume this option in this work. Any one of these
possibilities would lead to differentiated forms of identifying the target-population of the
study.
Another issue to be dealt with is the dating of retirement. Despite the fact that many authors
consider the individual's transition to retirement to be a process (in the sense that a
transition takes some time to be prepared by individuals and, thus, it is not reducible to a
point in time), we adopted in the paper a more “workable” definition of retirement, based
on what each adult respondent declares at a given point in time regarding his/her labour
market status, just like Bardasi, Jenkins, and Rigg (2002) do. Retirement is then considered
to take place in the first year the individual declares that being retired is her/his labour
market status.
10