14 / 22
14(21)
towards rehabilitation programs and SI rules respectively. If not entirely
random, matching seems close to random, displaying significant differences
in about half of the parameters presented. The differences are small
throughout, showing only weak signs of systematic matching between
individuals and caseworkers.
In the analysis, covariates are introduced in steps. The first model
(Model 1) analyzes the simple relationship between caseworker
attitudes and return to work without any controls. The second model
(Model 2) introduces local office (LO) fixed effects, taking into account any
heterogeneity owed to the local labor market or the local office. In Model 3
we add rich information on each individual on sick leave. As argued above,
matching between caseworker and individual seems non-systematic overall,
which means that adding this information should have little impact on the
attitude estimates.
9
Model 4 includes caseworker characteristics. This
enables us to learn to what extent the attitudes are related to
caseworkers´ gender, age, tenure, and education.
4.1
Impact on return to work
Table 4 reports the results from estimations of caseworker attitudes on
the return to work at 3, 6, and 9 months. The nine-month estimation is
performed on a subsample of individuals starting a sick-leave episode
between January 1, 2010 and December 31, 2010 (49,676 individuals).
Return to work covers either full- or part-time work and the estimates refer
to the impact of a maximum difference in attitude between zero and one.
Overall, the results are stable for different specifications. This suggests that
the impacts of the attitudes are not sensitive to the SIA local office or the
local labor market or to the caseworkers´ characteristics. It also implies
that unobserved heterogeneity in the matching between individual and
caseworker is probably not a problem.
The results show that a positive attitude towards rehabilitation programs
has no impact on return to work up to three months. However, up to six
and nine months respectively, the impact is negative, corresponding to a
2-3 percent lower probability of having returned to work. To the extent that
a positive attitude towards rehabilitation is associated with the caseworker
taking more rehabilitation initiatives, these negative effects are expected in
previous research showing no or negative effects of rehabilitation programs
on return to work. Also, given the probability that participating in a
rehabilitation program increases with sickness absence duration, an
increasing negative effect is also expected.
Furthermore, a positive attitude towards the SI rules increases return to
work for all durations. The size of the impact decreases somewhat with
sickness absence duration but corresponds to more than 2 percent at nine
months. If a positive attitude towards the SI rules increases monitoring and
eligibility checks, previous research suggests a positive effect on return to
work. Also, if the degree of moral hazard is higher among those on short-
term sick leave than those on long-term sick leave, the scope of monitoring
initiatives should be higher earlier in the sick-leave episode, explaining the
diminishing trend of the impact.
9
Altonji et al (2005) argue that the selection on observables is informative about the importance
of selection on unobservables. This should especially hold in this case with the rich data
available containing information on both the individual’s health status and the labor market
position.