Quality of reported ages: A robust re-modification in Total Modified Whipple’s Index

In demographic research, the accuracy of the reported ages in surveys and censuses is a persistently important issue. The common indices developed and used to examine the quality of age data are Whipple, Myers, Bachi's, modified Whipple, and the total modified Whipple's index. The most commonly used and simplest to compute index is the original Whipple's index proposed by George Chandler Whipple. It is a summary measure used to check age heaping on ages ending with digits 0 and 5. The other summary index is the total modified Whipple index by Spoorenberg. A robust modification is proposed for the total modified Whipple index. This modification, based on the method of the original Whipple index for all digits (0, 1, 2 … 9), is simple, robust, and easy to interpret. The proposed index is used to check the quality of age data from the Demographic and Health Survey data series of three countries (India, Pakistan, and Turkey). The original Whipple index only measures the preferences or avoidances of ages ending in digits 0 and 5. The existing summary index total modified Whipple index, based on modified Whipple indexes, covers all ages; however, results showed that the digit indexes for the terminal digits 1, 2, 3 and 4 have some added probability of being included. The newly proposed summary index, Robust Modified Whipple Index, is based on Rogers Whipple indexes, which display an equal probability for all ten digits i = 1,2,3…,9. In the application to the series of datasets, the proposed index is found better and more robust. The comparison of the results of the total modified Whipple index and proposed robust modified Whipple index concludes that the proposed modification is a more precise and robust to measure age misreporting by taking the effects of all terminal digits of reported ages. The proposed modification is suitable for the evaluation of the quality of self reported and interviewer recorded ages at the time of surveys and censuses.