A Comparative Study to Estimate and Forecasting Mortality Using Demographic and Statistical Models

Abstract

Estimation and forecasting for indicators of mortality are of great importance for the decision-making, it is considered to be the main economic and social planning factor which contributes to the development of policies and Strategies in the long term. The problem of this study is that, there is no studies on mortality and life expectation at birth (e0) as an indicator of mortality levels, the studies on life expectation at birth (e0) are scarce in Sudan due to the problem of incomplete vital records of mortality beside, the high rates of illiteracy along together with some other reasons have contributed to the problem of the incompletion of the records .The importance of this study is that, there is no study on estimating and forecasting life expectation at birth (e0) using demographic and statistical models in Sudan because such study needs accurate data therefore, this study will be of great benefit in the areas of statistics and its applications. The study aimed to apply statistical and demographic models used in the field of mortality and adapted to be use in Sudan for estimating and forecasting life expectancy at birth (e0), also the attempt to correct defective data in mortality to study the phenomenon understudy in Sudan using demographic and statistical models(and compare between them). The use of population censuses through indirect estimation techniques, to determine levels, trends and patterns of mortality in Sudan .The hypotheses of the study are: It is mathematically plausible to derive an adjusted model life table to estimate the mortality in Sudan. The time series of life expectation at birth (e0) in Sudan for period (1973 – 2008) is stationary. There is no significant difference between statistical and demographic models to forecast the phenomenon understudy. The research methodology comprises of descriptive and inferential statistical methods also used demographic techniques. The data were obtained from Sudan Population Censuses (1973 -1983- 1993- 2008)., using programs (Statistical Package for Social Sciences, Excel, Eviews and Spectrum) in order to reach results. The study found that, the levels of mortality in Sudan are high compared to developed countries. Life expectation at birth (eo) increased by about 14 years for both sexes, 13 years for males and 12 years for females during the period (1973-2008). When we look to the pattern of deaths in Sudan at that time, we found that, the central death rate (nMx) by age groups took the form of the Latin letter (U). The study also used demographic models and (Box - Jenkes time series -ARIMA) as statistical model to forecast the life expectation at birth (eo) during the period (2009-2020) for each of the (both sexes - males - females). The study found that, the time series is not stationary for both sexes and males, but the series is stationary for females, in order to make time series stationary for both sexes by taking the second difference of the series, the best model to forecast (both sexes) is ARIMA (0,2,1), as for males it stabilized the series after taking the first difference and best model to (males) is ARIMA (1,1,0). For females the best forecasting model is ARMA (1, 0), .The study made comparison between demographic and statistical models, which is better in forecasting the life expectation at birth (eo) for each of the (both sexes- males - females), as one of the objectives of the study. The study concluded that, the Statistical model is better than demographic model when data are homogeneous, used in several statistical criteria’s to compare between statistical and demographic models. This study affirmed the importance of demographic indicators reached. It is necessary to develop demographic package programs such as (MORTPAK Package and CHADMOR Package) in order to apply the modern demographic models in a country like Sudan suffering from deficient and incomplete data problems. The necessity to increase the volume of data and remove all the discrepancies of time series data before the application in order to raise the efficiency of each of the models Box –Jankins and other demographic models in life expectation at birth (e0) analysis. Separate data studies should be performed when the data are homogeneous and another study when the data are heterogeneous focuses on the comparison of statistical and demographic models to forecast life at birth (e0) using other models not used in this study.