13 / 22
13(21)
4
Analysis
To analyze the relationship between the caseworkers’ attitudes and
whether or not the individuals returned to work after different durations,
we estimate a linear probability model:
,
'
'
'
t
j
i
ij
j
t
ijt
CW Ind
LO
Index
Y
'
(1)
where
ijt
Y
is the binary outcome for individual
i
assigned caseworker
j
of
having returned to work at time
t
, represented by 3, 6, and 9 months
respectively.
j
Index
corresponds to the standardized index values of the
caseworker´s attitudes towards rehabilitation programs and SI rules.
ij
LO
is the local office at which the individual and the caseworker are registered,
and
i
Ind
and
j
CW
are vectors of individual and caseworker characteristics.
,
,
, and
are coefficients capturing the relationship between the
variables and the marginal probability of having returned to work.
For
to capture the causal relationship between the caseworker’s
attitudes and the individual´s return to work, the assignment of individuals
with different expected sickness absence duration must be independent of
the caseworker's attitudes after conditioning on
Ind
. If, for instance,
individuals with worse health and/or lower work motivation systematically
are assigned caseworkers positive to rehabilitation programs, the estimate
will be biased downwards if available data do not fully capture health status
and work motivation. However, since the caseworkers´ attitudes should
be unknown to anyone but themselves, there is no obvious process in
which individuals with different expected sickness absence duration could
be matched with caseworkers with particular sets of attitudes. This is
especially expected to be true regarding unobserved individual
characteristics.
Systematic matching between caseworker and individual characteristics
could, however, arise if caseworkers are exposed to different types of
workers. To find out more about the assignment routines, a survey among
representatives at all SIA local offices was performed in 2010 (ISF, 2014).
In general, three criteria were applied in the allocation of individuals to
caseworkers: employment status (employed/not employed), employer
(sector), and caseload. Also, more than half of the local offices allocated
individuals according to which day of the month they were born, at least
among subgroups. None of the offices reported allocation based on the
individual´s expected sick-leave length.
From the data, we get an idea of the matching between caseworkers and
individuals in practice. In Table A1, we compare the individuals assigned
a caseworker above and below the median caseworker regarding attitude