28
6
Effects for Unemployed and Employed
We repeat the above analysis in the Sassam experiment for each month and state, now
separately for the employed and unemployed.
In Panel A of Table 6 we display the effects on sickness absence. From this panel we
can see large differences in effects between the two groups. For the employed we find
evidence of locking in effects in February 2008, but from March the effect is basically
zero. For the unemployed, the locking-in effect is larger but the effect is also high and
statistically significant until November. The relative risk of being sick absent if prioritized
is 37 percent higher in March than if not prioritized (53 percent of the controls are on
sickness benefits). In October the corresponding risk is 27 percent (35 percent of the
controls are on sickness benefits).
There are no statistically significant effects on unemployment among the employed
(Panel B). The point estimates are in general lower for unemployed, and for this group we
also have weakly statistically significant effects in some months.
From panel C we can see increasing risk of getting DB if being prioritized to Sassam for
both employed and unemployed. The effects are statistically significant from June and
forward. After the first 6 months, the impact is generally about twice as large for the
unemployed as for the employed. For example, the uptake in June for the unemployed
controls is 3 percent compared with 1.5 percent among the employed controls. In
February 2009, the corresponding figures are 7.5 percent and 4 percent. This means that
the differences in effects are also highly economically relevant. Being prioritized to
Sassam would increase the likelihood of having DB from 7.5 percent to 10.6 percent 15
months later. The corresponding figure for the employed is an increase from 4 percent to
5 percent.
The result from this analysis supports the hypothesis that the effect observed in
section 4 can stem from the lower cost of communicating or signaling health problems
when being assessed in a Sassam meeting. Parsons (1978) suggests that rehabilitation
initiatives among the sick absent individuals could create or magnify a “sickness identity”
of individuals making use of sickness or disability insurance. This could then lead to an
increased take-up of sickness and disability benefits, instead of decreased. If unemployed
individuals in comparison with the employed are more likely to get a sickness identity
when being assessed or exposed to an intervention, this would generate the same pattern
as observed in our data, and then the suggested analysis cannot discriminate between
the two theories.
