Table 3

Studies on Swedish and Norwegian policy changes

AuthorPopulationStudy typeType of policy variation and disability benefit scheme investigatedResult: regression coefficient (p value)CommentsVA
Hesselius and Persson (2007)34All individuals with sickness absence spells of at least 91 daysDifferences-in-differences approach using panel dataVariation in benefit generosity only. National sickness insurance 1998Linear regression of policy change on duration of sickness absence 4.66 days (p=0.001)The authors concluded that this policy resulted in an increase in the duration of sickness absence, in this population by an average of 4.7 days. No corresponding effect was found before the 91st day or after the 360th day in sickness absence. Changes in health or separate time trends in educational or occupational groups are not adjusted for in the analysis.14
Karlström, et al (2008)43Male workers aged 60–64 yearsDifferences-in-differences approach using panel dataVariation in eligibility requirements only. National disability insurance programme, 1997Linear regression of various transitions in and out of employment
Employment: non-employment
−0.0074 (p>0.1)
All states: disability insurance −0.0104 (p>0.1)
non-employment→ non-employment 0.01(p<0.05)
The authors concluded that it was not possible to detect any effect on employment from the reform. There did, however, appear to be an anticipation effect, in that there was an increased flow into disability insurance when the reform was announced. This was 2 years before the reform was actually implemented.
They did, however, find that the reform was associated with a decrease in transition from unemployment insurance to disability insurance, and higher transition from employment to sickness insurance and lower transition from sickness insurance to disability insurance as well as increased persistence in sickness insurance. In other words, the reform resulted in people shifting between benefit systems and not into the labour market. Changes in the level of disability, wages and/or benefit level were not controlled for in the analysis
Bowitz (1997)35Men and women 16–66 yearsA time series approach using ecological dataVariation in benefit generosity only. Disability pension 1971 to 1991An error correction weighted linear regression analysing the effect of the replacement rate on the probability of entry into disability benefits 0.17 (p=0.16)The authors concluded that unemployment was important in explaining rising entry rates into disability benefit, but that there was less evidence for the effect of increases in the replacement rate. No control for changing health status or educational level.9
  • VA, validity assessment.