10(30)
We observe the parents when they are in their 40s and the children when
they are between 27 and 36 years old.
13
The information about sick leave is
added to our register information on the total population.
To receive information about the link between children and their parents-in-
law, we proceed as follows. In the registers, we can observe which persons
share households. In order to identify couples – or two adults who live
together – we first exclude all the household members below 17 years old
and persons who are above 16 years old but who share a household with
their parents. Second, we eliminate all single households. Finally, we only
keep households with two adults of different sex.
14
4.3
Definition of sickness absence
Sick leave in this study is defined as days with sickness benefit and days
with disability pension reimbursed by the SSA.
15
In order to measure sick
leave in the same way over time, we start counting days from the fourth
week within a given sickness spell. With respect to disability pension, we
observe the first day replaced and start counting from the first day in a
sickness spell with a disability pension. The annual number of days on sick
leave could sum up to more than 365 days per year. The maximal number
of days was set to be 365. (This concerns only around 1 per cent of the
individuals in the study population; see Table 1.)
Individuals could be entitled to sickness benefits on a full-time or a part-
time basis. Due to data restrictions, we focus on the number of days with
sickness benefits rather than the extent of this type of benefit. When it
comes to disability pensions, we make a distinction between a disability
pension on a full-time basis and a disability pension on a part-time basis.
16
As a robustness test, we re-estimate all our models with alternative
measures of sickness absence, such as annual prevalence of sickness
absence (1/0) and the annual average number of sickness absence spells.
The qualitative conclusions are the same irrespective of which
measurement we use.
In order to obtain a representative picture of the level of sickness absence,
we observe both parents and children during a period of six years. By doing
so, we avoid occasional fluctuations in sickness absence that are not
13
Since sickness absence varies across ages, we (if anything) estimate a lower
bound of the true intergenerational correlation when parents and children are
observed at different ages. In the analysis, we control for both the parent’s and the
child’s age with a second-degree polynomial. The distribution of ages of parents
and children is large; hence, we believe that our estimations are unbiased.
14
Adult persons who live with a parent or who live with a sibling cannot be excluded.
After our restrictions, such cases should be extremely rare; hence, we believe this
part of the analysis is mainly based on adult couples.
15
The reason for including both sickness benefit and disability pension is that a long
period of sickness benefit in many cases ends up in a disability pension on a more
permanent basis. Thus, by only studying persons with sickness benefit, we miss
the persons who are the most absent due to sickness.
16
For the parents with respect to disability pensions, we only have information about
the status at the end of December each year. We multiply this information by 365
days to obtain an approximation of the annual average. For the children with a
disability pension, we have information on a daily basis.
