Sickness absence and voluntary employer paid health insurance

Sickness absence is a problem with considerable economic dimensions. About 4% of the total annual working days are lost due to absence. Therefore, ways to reduce absence are eagerly sought. In a Danish context employer paid insurance is but one example. The tax exempt status of this type of voluntary duplicate health insurance has been argued by reference to the potential for reducing long term sickness absence. However, nationally and internationally there is no evidence about this. The present paper analyzes this theoretically and empirically. A simple model for ‘demand’ for sickness absence in the Grossman-tradition is used. Empirically, two recent survey data sets are used. The determinants of absence are analyzed using quantile regression in order to look at the extreme parts of the conditional distribution, e.g. 90% og 95% for long term absence. No significant results are found on the absence reducing property of health insurance. A two component (‘short’ and ‘long’ term absence) finite mixture model is also applied with the same result. The problems with a causal interpretation of regression analyses may (partly) be circumvented by using (correctly specified) propensity scores and matching estimators. Regression analysis and propensity score, however, share the same challenge: Both are based on selection based on observables. Using the matching estimator approach there are no signs of a treatment effect of health insurance using the presenteeism data set, while there is evidence using the health insurance data set. However, the specification of the propensity score for the latter is not as exhaustive as for presenteeism data set, and in some cases there are statistically significant differences for some control variables after matching.