We consider the exit routes of older employees out of employment around retirement age. Our administrative data are high dimensional as they cover weekly information about the Danish population from 2004 to 2016 and 397 variables from 16 linked administrative registers, covering a wide range of information such as demographic, socioeconomic, financial, criminal, labor and health information, etc. We use a flexible dependent competing risks quantile regression model to identify how exits to retirement, illness, unemployment, etc. are related to the information in the various registers. To help finding an appropriate model we use variants of the Lasso, in particular the (adaptive) group bridge applied to competing risks quantile regression model to identify the relevant administrative registers and within-register variables. To our knowledge, this is the first application of these methods to large scale administrative data and the problem of exit into retirement. It is found that selected registers and most within-register variables from the (adaptive) group bridge have reasonable interpretation and remain significant in the unpenalized competing risks quantile regression. By applying state-of-the-art statistical methods to large scale data, we obtain detailed insights into the conditional distribution of transitions from employment into retirement in the presence of high dimensional data and competing risks setting.
|Educations||MSc in Advanced Economics and Finance, (Graduate Programme) Final Thesis|
|Number of pages||71|