31
of threat effects also
24
in the sickness insurance (cf. Hesselius et al., 2005; Hägglund,
2010; Johansson and Lindahl, forthcoming; de Jong et al., 2011). That is, an increased
hazard from sickness absence when individuals are monitored and before entering
rehabilitation schemes.
25
de Jong et al. (2011) also find a reduction in applications to DB.
To further our understanding we formulate a theoretical model in section 5. In this
model sick absent individuals may want to receive DB (secure payment and no work). In
order to get DB an individual communicates or signals health problems when being sick
absent. There is, however, a cost of signaling, and this cost is reduced if the individual is
being assessed by the SIA. The implication is hence that the treated individuals who
prefer work absence in front of work remain sick absent longer and then obtain DB to a
higher degree than those not assessed. The model also predicts that the effect from being
assessed should be larger for those with bad health and for those with low costs of being
work absent (or low incentives to resume from work absence). This hypothesis is then
tested in section 6, where we estimate the effects separately for unemployed (low
incentives to resume and/or bad health) and employed. We find that the effects on
sickness and disability benefits are larger for unemployed than for those who are
employed at the start of their sickness absence spell, which was expected from our
theoretical framework. This, hence, does not reject our hypothesis of a signaling effect.
In the literature it has been hypothesized (e.g., Parsons, 1978) that rehabilitation
could give the sick absent individual an identity as being ill, which then will cause longer
and/or more sickness absence. Based on Swedish data, Anderzén et al. (2008) also found
that active rehabilitation could prolong, rather than shorten, sickness absence, which
hence supports this theory. The theory discussed above would be an alternative. If the
unemployed individuals, compared to the employed, are more likely to define themselves
as ill when being assessed and/or exposed to vocational rehabilitation, then the two
theories cannot be tested using the analysis performed here. The theoretical framework is
to some extent related, or rather the opposite, to the self-screening model suggested by
Parsons (1991). In this model long waiting times to decision will lead to fewer claims of
DB. The reason for this is that there is uncertainty of a decision from a claim and there is
cost of waiting and not working if the claims are not granted. Furthermore, if the time
preferences for individuals with good and bad health are the same, the cost of waiting for
those with good health is larger than for those with bad health.
24
In the unemployment insurance literature there is, by now, plenty of evidence of increased exit
rates from the unemployment insurance before monitoring and screening and program start (see,
e.g., Black et al. (2003), Geerdsen (2006), Geerdsen and Holm (2007), Rosholm and Svarer
(2008), Graversen and Van Ours (2008), Arni et al. (2009), and Hägglund (2011)). These effects
are known as threat and pre-program (or pre-treatment) effects in this literature.
25
Somewhat related, Borghans et al. (2010) find that an increased stringency in the Dutch disability
insurance increased the outflow from disability insurance to social assistance.
