In recent years, there has been an increasing interest in broadcast database systems, where the server periodically broadcasts contents of a database to mobile clients such as PDAs. There are three query processing methods in the broadcast database system: (i) the server processes a query and then broadcasts the query result to the client; (ii) the client stores all data that are necessary in processing the query and then processes it locally; and (iii) the server and the client collaborate in processing the query. Since the performance of each method changes according to the system situation, such as query frequency, it is difficult to decide the optimal method among them statically. In this paper, we propose a new query processing method which dynamically changes the query processing method based on query frequency. This method improves not only the response time but also the success rate of query processing compared with the traditional methods.