Death Registration:An Assessment
3.2. Data and methodology
3.2.4. Criteria for assessing the four techniques and method of presenting results
Given that each technique has a unique assumption (except Brass’s and Preston and Coale’s methods) and some differing strengths and weaknesses, assessing the applicability of these techniques to Philippine data, and their robustness and consistency, falls under several conditions. First, if the population under study is stable and age reporting is accurate, Brass’s and Preston and Coale’s procedures are likely to yield consistent and plausible levels of completeness of death registration. If there is age-misreporting, Brass’s method is preferable, owing to its robustness to age*misreporting. Second, if the population in question is not stable, whether or not there is age-misreporting, which is likely to characterize data in developing countries, the Gray, Preston and Coale and Brass approaches are likely to reveal consistent results. Most often, results obtained with the Gray’s and Preston and Coale’s techniques or with the Gray and Brass procedures, show close correspondence, in view of the nature of their assumptions and strengths as discussed above. In most instances, the Gray approach is the most robust because it assumes non-stability and is less sensitive to age-misreporting; the Preston and Coale method may rank second for it is fairly robust to violations of stability but more sensitive to age-misreporting; the Brass technique may rank third if violation of the stability assumption is more crucial than violation of accurate age-reporting. If the reverse holds true, then the Brass approach would be preferable to the Preston and Coale procedure but not to the Gray technique. It is the extent of violations of their assumptions that leads to these three techniques yielding different results.
Third, in populations where internal migration is heavy, the Courbage and Fargues approach is the most reasonable source provided the assumed model is a close representation of the actual mortality of such populations. In non-closed populations, all the Gray, Brass and Preston and Coale approaches are likely to produce results less meaningful than the Courbage and Fargues approach. As the Courbage and Fargues approach does not assume a closed stable population, the appropriate choice of a model is important for its application. In fact, if the chosen model is an appropriate representation of the mortality of the population in question, results obtained with this technique are likely to be consistent with one, two or all of the other three approaches depending on the prevailing demographic change in a given population.
Owing to the unique assumptions and differing strengths and weaknesses between or among the four techniques, there will be no, or only occasional, close correspondence between the results obtained using all of them. The likely situation, when all four show consistent levels of completeness of death registration, is when there are very few registered deaths, causing the application of all four techniques to reveal equivalent low values, which may be mainly statistical artefacts.
It can happen that the application of an indirect technique, no matter how powerful it is, does not help. In such situations, however, a remedy is to examine the types of errors besetting the data, and then adjust the data for these types of errors before attempting to apply any indirect technique of estimation.
In determining the extent to which these four methods are consistent with each other, several pairs or combinations of them evolve from the above four conditions. Only eight of the most possible pairs or combinations are taken and are summarized below:
Condition Pair or combination
Stable closed population with or without age-misreporting Non-stable closed population
with or without age-mi sreporting
Appropriate model mortality pattern with:
stable open population with or without
age-mi s repo rting
non-stable open population with or without
age-mi sreporting
Extremely poor registration of deaths
Brass's and Preston and Coale's
Gray's and Preston and Coale's Gray's and Brass's
Gray's, Brass's and Preston and Coale's
Brass's and Courbage and Fargues's
Gray's and Courbage and Fargues's Preston and Coale's and Courbage and Fargues's
A combination of all four methods
One can establish many levels of consistency for any given pair or combination. However, in this study, three levels are arbitrarily taken and should be treated as a special not a technical term throughout the thesis. The first is close congruence (CC), arbitrarily set to a difference of less than or equal to 10 percentage points between the highest and lowest values yielded by the techniques being compared. Second is moderate congruence (MC), which means a difference of 11 to 20 percentage points in the estimated level of completeness yielded by the techniques under comparison. Third is poor congruence (PC), which refers to a difference of over 20 percentage points in the level estimated by the techniques paired or combined. The respective variants of the Gray, Preston and Coale, and Courbage-Fargues approaches are not treated singly; the lowest or highest among the results of the variants of each method is chosen in the consistency check to represent the method.
If close or moderate congruence is evident in at least one of the above eight pairs or combinations adopted in the present assessment of these four procedures, it may be taken that the data are good; hence reasonable levels of death recording can be derived. Presence of poor congruence implies strong violations of the assumptions of the relevant procedures or seriously inaccurate data. If the paired or combined procedures are Gray’s, Brass’s and Preston and Coale’s, the estimated relative variances provide a basis as to which procedure is more or the most robust
A discussion of the results of assessing the four methods covers all areal aggregations and periods under consideration. These discussions mainly refer to the tables in the appendices.
3.2.5. Criteria for choosing the best estimate of the level of death registration completeness