Myers’ Blended Method: An Alternative Approach

Abstract

Myers’ index is widely used in censuses, large scale sample surveys and in many other secondary data to study the digit preference error on single year raw age distribution particularly in developing countries where huge peaks and troughs are observed at the ten digits of age. Historically many modifications are known and theories have been developed on this method by actuaries, demographers and other social scientists (Bachi, [3]; Carrier, [4]; Ramachandran, [5]; Mukhopadhyay, [6] and others). Usually the well known Myers’ Blended Method (1940) is done where the figures for ages within a broad band on ten digits from zero to nine are much significant and high. The same method has been tried in the paper to find the quality of age reporting in a small area sample in the district of Murshidabad, West Bengal, India. The index value seems to overestimate may be because of too small values for some of figures corresponding to the ten digits of the entire distribution even within a broad age range of population. An alternative approach replacing the original population by person years is made in the paper to calculate the same index. The modified index value seems to be appropriate keeping in view of the type of high and low values only for a few digits out of the ten digital values.