The anonymisation of linked employer employee datasets constitutes a special problem for data producers. Concerning the employees there is generally less risk of reidentification, but their data can be used to identify the employer. We present a strategy that permits to measure dependencies between employer and employee data, to evaluate whether these dependencies have an impact on the reidentification risk of the employer and, if necessary, to anonymise the data of the employees in such a manner that the confidentiality of the employer is guaranteed. We apply this strategy in the generation of a scientific use file of the German Structure of Earnings Survey 2001.