31
Table 9: Heterogeneous effects on inpatient care with respect to pre-retirement income
(column 1) and education (column 2) during ages 56-70
Income Education
Main effect
-0.9779**
(0.328)
-1.0375**
(0.3726)
Heterogenous effect with respect to income
0.0809*
(0.0315)
Heterogenous effect with respect to education
0.7022*
(0.3534)
Evaluated at p25 (231,000 SEK; 26,837 Euro)
&
-0.558
-
Evaluated at mean (295,000 SEK; 34,273 Euro)
&
-0.442
-
Evaluated at p75 (335,000 SEK; 38,920 Euro)
&
-0.369
-
Evaluated at
educ
= 1 (high education, college)
#
-
-0.3353
Note: Estimation is performed with the Poisson maximum likelihood estimator. Robust standard errors in ():† p<.1; *
p<.05; ** p<.01. &The evaluation is based on the specification
̂
̂
, where inc54 is income at age 54.
#The evaluation is based on the specification
̂
̂
, where educ is 1 if college educated. The number of
observations is 19,986. All models control for the main effects, that is, inc54, educ, military dummy and dummy for
cohort 1938-1939.
6.5
Sensitivity analyses using alternative morbidity outcomes
Based on our detailed register data, we can define an (almost) infinite number of
outcome variables. To this end is, thus, the choice of number of days in inpatient care an
arbitrary measure of a health outcome. Aside from testing for selective censoring due to
death, the analysis of mortality provides a sensitivity analysis for the results obtained
using inpatient care outcomes. However, for completeness we also estimated the same
type of regressions as with the number of days using the number of episodes (visits that
meant inpatient care), and annual prevalence of inpatient care (1 if at least one visit
during a year and 0 else) as well.
In all these regressions, receipt of the offer of early retirement decreases both the
yearly prevalence of inpatient care and inpatient care visits, and all the estimates are
robust to the inclusion of control variables.
27
These results therefore confirm the
previous results, using the number of days in inpatient care only.
7
The effects of retirement on health
Earlier studies – discussed in the introduction – have focused on the effects of
retirement on health. We argued that cross-sectional analyses could be questioned, as
individual decisions to retire are most likely influenced by health reasons. That is, the
population sector that retires early has worse health overall than the population sector
27
The results can be obtained from the authors upon request.
