Childhood Mortality Estimation

Preliminaries
The results of this section should be read bearing the following considerations in mind:

1. When estimating mortality levels we need a so-called mortality model. We have chosen the West model life table (Coale and Demeny 1983) because it has been shown by the UNICEF and JFPPA regional office to fit Middle East populations reasonably well. Moreover, in the absence of literature on mortality trends in the region, the West model, being an average model based on a large number of empirical life tables, is a primary choice.
2. The methods being used to estimate q(1) and q(5) usually produce better results for q(5) than for q(1). Hence, one should always treat estimates of q(1) with caution.
3. Estimates of q(x), x=1,5, based on the first and to some extent the second age groups (or duration groups), are normally discarded on account of their instability.
4. In the case of this data set, a single estimate of q(x), x=1,5, will be produced by calculating the weighted average of q(x) for the second and the third groups with weights being the proportion of CEB in each group.
5. Trends of q(x), x=1,2, will be presented in common indices to facilitate comparisons.
6. The data are weighted and the effect of weighing can be seen from the CEB values since they are given with decimals and not as integers.

Estimates Using Data Classified by Age Group of Mothers
Table 2.6 presents input data for analysis. Disregarding gender, the above table shows that 9%1 of reported CEB have survived by the time of field work. Using data of the above table and the methodology described in UN (1983), the calculated mean age at child- bearing is 26.35 years for both sexes, 27.91 years females, and 25.84 years for males.

Table 2.6 Number of children ever born and children surviving by age group of mothers and sex of children

Further calculations show that the probability of dying before age x is almost identical for the two standard alternative procedures (Palloni-Heligman, Trussel). The only slight difference is in the estimated reference dates. However, the differences between the reference dates are not of a large magnitude. For trend assessment, estimates of q(1) and q(5) can be converted to common indices as shown in table 2.7 together with life expectancy at birth, e₀.

Table 2.7 Estimates of probability of dying by exact age 1 an d 5 and life expectancy at birth (in common indices) using two different versions

Inspection of the estimates of q(1) and q(5) shows a clear mortality decline during the last 15 years. The high estimate of q(x) for May 1990 is expected since it is based on an age group (20-24) where many women marry and have their first child, there is usually a surplus mortality for first births.
Life expectancy at birth is calculated using values of q(1), which explains the upward trend in life expectancy.

To get a single estimate for q(1), q(5) and e0, we use the weighted average of the estimates of each parameters for age groups 20-24 and 25-29. The weights suggested here are the proportion of CEB in each age group (0.336 for 20-24 and 0.664 for 25-29). This is done to avoid discarding the mortality experience for the 20-24 age group altogether, and to obtain estimates that are fairly recent.

Our estimate of q(1) is 43 deaths per one thousand live births for both the Palloni-Heligman and Trussell approaches. The Palloni-Heligman estimate refers to February 1989 and the Trussell's estimate refers to December 1988, both of which in practice refer to the same time period. The estimate of q(5) for the same reference dates is 56 deaths per one thousand live births. The two procedures also give very similar estimates of q(1) and q(5) for each sex.

We have also estimated child mortality using children ever born and children surviving by duration of marriage of mothers, using Trussell's procedure, see table 2.8.

Table 2.8 Estimates of probability of dying by exact age x, 1 and 5 and life expectancy at birth, Trussell version

Generally, estimates of q(x) based on data by duration of marriage are expected to be more reliable than those derived from data by age of mother, due to a higher accuracy in reporting age at first marriage (being a memorable experience). However, for this data set, the sample size is not large enough to permit reliable estimation. Therefore, since we are looking for trends in IMR and U5MR rather than specific point estimates, both approaches show a decline in q(1) and q(5) during the last fifteen years, but the decline based on the age-of-mother data is more systematic. Moreover, the decline in q(1) and q(5) resulting from the age group method agree quite well with that reported by UNICEF and JFPPA for the same period. Based on these arguments, we recommend adopting the q(x) values based on data by age of mother.

al@mashriq                       960715