Longevity and mortality risks are now threats to those who are undertaking pen-sions such as governments and insurance companies. Therefore, mortality forecasting is playing an increasing important role. The Lee-Carter Model, the American offi-cial method, is the most popular among all the researches, especially in the dynamic models.The paper is mainly about the dynamic modeling for kt, which is the Lee-Carter decomposition result in connection with time. The key point is the consideration about the jump points. It is worth mentioning that this paper is the first one to use double-exponential model on Hong Kong population. There are three cases including the one without jumps, the one with single exponential jumps and the one with double-exponential jumps. Using moment estimation, we forecast the HK mortality rates and compare the results. The conclusion is that the improved Lee-Carter Model is suitable for HK population, that error estimation is larger for older ages and the special cases are needed for older ages, that the models with jumps have better forecasting results in long-term forecasting. |