| The purpose of this paper is to assess the performances of the existing classic model life table systems in applying to China and to construct a new model life table system for China. The text is divided into 4 chapters. In first chapter that is research background, we mainly raise the probable problems on the application of the existing model life table systems in China, which is whether the existing model life table systems can reflect the Chinese age-specific mortality rate. The comprehensive review of the model life table systems was provided in the second Chapter, where we introduce their development and current situation. There are nine major types of empirical models which are well-known and widely used:(1) Valaoras (1955); (2) Lederman (1959/1969); (3) Coale & Demeny (1966/1983); (4) Brass relation model (1971); (5)United Nations mode life table for developing countries (1982); (6) Murray (2003); (7) Wilmoth (2012); (8) Clark and Sharrow (2012); (9)region model life table for China(Jiang,1990).We also introduce the methods for generating these model life table systems in this chapter. Chapter 3 assesses the performances of the existing model life table systems in applying to China. Chapter 4 constructs a new model life table system for China and accesses its performances of application in China. The details for the last two Chapters are as follows:Chapter 3 the Assessment of the Existing model life table systems in applying to ChinaObjective:To evaluate the performances of five most widely used model life table systems: Coale-Demeny model life table, United Nations mode life table for developing countries, Murray model life table, Wilmoth model life table and Clark-Sharrow model life table. To investigate whether these five systems can capture Chinese age-specific mortality.Methods:We divided into two parts to assess their performances according to their different situations of the application for these 5 systems. The first part is for the latest three systems:Murray model life table, Wilmoth model life table and Clark-Sharrow model life table. We collect 19 years life tables by sex for each different population between the periods 1994-2012 in China. These three systems were used respectively to estimate complete life tables for those life tables of our collections manily considering the under-five mortality rate (sqo) as an input parameter. Then, the estimation errors between these actual life tables and model life tables for four key mortality indexes will be investigated from three aspects:the life expectancy at birth (eo), the infant mortality rate (1qo), the adult mortality rate (45q15) and the old age mortality rate (20q60). First, we analyze the residual errors between estimated and observed value for these four indexes and investigate whether the system biases exist. Second, a deep comparison was made respectively between different population and different gender. Third, a further comparison was made between China and other 12 countries.The second part is for Coale-Demeny model life table and United Nations model life table for developing countries. We analyse the relationship between 5qo and the four indexes:eo, 1qo,45q15 and 20q60 respectively and investigate whether the Chinese observed value fall outside the range estimated by these nine mortality patterns of these two model life table systems.Results:For the first part for the latest three model life tables:first, the plots of residual errors for four mortality indexes show that almost all the residual profiles for each sex are either below or above the horizontal line of residual 0. For example, the residual profiles are below the horizontal line of residual 0 in most cases for eo and are below this line for 20q60. Secondly, the mean relative errors (MREs) for female are larger than for male for e0,45q15 and 20q60, but exception for 1q0. Moreover, the MREs for urban population are more than for rural population. Thirdly, the MREs for China are more than those for other 12 countries in most cases.For the second part for the other two mode life tables:No mater for which model life table system, the Chinese observed value is inside the range estimated by the model for high mortality levels, but for low mortality levels, most of the Chinese observed value fall outside this range. On the other hand, the best mortality pattern for China is the West pattern for Coale-Demeny model life table and is the Far Eastern pattern for United Nations mode life table for developing countries on average.Conclusion:First, Murray model life table, Wilmoth model life table and Clark-Sharrow model life table cannot capture Chinese age-specific mortality and its application exist a significant system bias. For example, the eos are also underestimated in most cases using 5q0 as an input parameter. Secondly, Coale-Demeny model life table and United Nations mode life table also cannot fully reflect Chinese age-specific mortality, especially for the low mortality levels. But comparing with other three model life table systems, the e0s are overestimated.Chapter 4 Model Life Table for ChinaObjective:To construct a new model life table system for China and then evaluate the errors in applying to China, comparing with the existing model life table systems.Methods:We use the region-specific dataset (provinces, municipalities or autonomous region) of the year 1990,1995,2000,2005 and 2010. The data from 1995 and 2005 is the data of a nationwide sample and investigation. The data from 1990,2000 and 2010 is respectively from three censuses. We corrected some age-specific mortality rates for the eld-old ages (over 80 ages) and the ages with a lower death number using three different methods:Coale-Guo method, Kannisto model and Brass-logit model according to the different characteristics of these five years dataset. The method we use to construct the new model life table for China is the modified Brass-logit model by Murrary. The corrected data was used to fit this model. The new model life table for China could be generated in different input cases by procedure of software R. The measure to evaluate the new model life table for China is the root-mean-squared error(RMSE). We compare the RMSEs with the existing model life tables.Results:The new model life table for China is a two-parameter system. Comparing the performance of this new model with the existing model life tables in applying to China, the RMSEs for this model are less than for other existing models for eo, iqo of female,45q15 and 20q60-For RMSEs of 1q0 for male, the new model life table is slightly more than Wilmoth model.Conclusion:The new model life table for China had already been constructed. The new model performs better than other existing model life tables in the application in China and do not exist system bias. |
PDF Full Text Request