9(30)
4
Empirical strategy and data
In this section, we present the empirical strategy, the data used, our
definition of long-term sickness and the study population.
4.1
Empirical strategy
The empirical strategy in this study stems from theoretical ideas about
intergenerational mobility presented by Becker and Tomes (1979, 1986).
The basic idea is that parents affect their children both through investments
in the child’s human capital and through the inheritance of ―endowments‖.
The Becker–Tomes model has been used in empirical studies of
intergenerational mobility with the aim of capturing the total inheritance
passed from parents to their children (see, for example, Solon, 1992; Solon
et al., 1991).
In empirical studies of intergenerational mobility inspired by the Becker–
Tomes model, an explicit aim is to observe the parents in a representative
period of their lives and during a period when their children may be
affected by their actions. We follow earlier empirical works on
intergenerational mobility (see, for example, Björklund, Roine and
Waldenström, 2012) and observe the parents when they are in midlife and
when they have children of school age. The children are observed when
they are about the same age as their parents are when they are
observed.
12
4.2
Data
The analysis is based on register data covering all Swedish citizens aged
16–65 years. The parents are observed between 1986 and 1991 when
their children are between 10 and 19 years old. The children are observed
between 2003 and 2008. We have annual register information on
employment status, earnings and household units, i.e., we can link family
members to each other. We can also link all children to their parents. In
Table A1 in the Appendix, we present the structure of attrition due to the
merging process of different registers.
We have information on sickness absence from population-wide registers
provided by the Swedish Social Insurance Agency (SIA). These registers
cover all days with sickness benefit and disability pension for all Swedes
since 1986.
12
Due to limitations in the available register data, we are not able to fulfil this
criterion perfectly. In our analysis, the parents are in general about 5–10 years
older than their children when observed.
