Linking survey data to administrative data can increase the possibilities of analysis. Because of its high accuracy exact linkage is the preferred linkage method. However, the linkage process involves several problems and typically only a percentage of records can be exactly linked. Sometimes this percentage can be quite low if non-consent rates are high or if the unique identifiers used in the linkage process are of poor quality. The lack of complete linkage can reduce the quality of the combined data and inferences that are drawn from them.
Thus, we evaluate the use of statistical matching to link records that cannot be exactly linked using unique identifiers. The records which could be exactly linked are used as the "gold standard" against which we evaluate the quality of statistical matching procedures and the resulting inferences.
In this project we use the German Panel Study "Labour Market and Social Security" linked to the "Integrated Employment Biographies" (IEB) data. The IEB is an administrative database that contains observations of employment history, benefit recipient history and job search on a daily basis collected by the Federal Employment Agency.
|