The KERIS Union Catalog is the largest bibliographic database in Korea. It has over 5.7 million bibliographic records and at present 330 university libraries are participating in shared cataloging services. The database, however, includes various errors and a large number of duplicate data because separate databases from many university libraries were merged without proper quality control in a short period. This study intends to find error data types by analyzing dissertation bibliographic data. The results show that error data are mainly caused by input errors, and the mistakes in using MARC formats and cataloging rules.