Assessing the reliability of the 1986 and 1996 Lesotho census data

Abstract

The present study attempts to assess the quality of reported age–sex distributions of Lesotho's 1986 and 1996 censuses using conventional demographic techniques. First, the data presented in single years are examined to identify patterns of digit preference and then Whipple's, Myers' and Bachi's indexes are computed in order to furnish a clear picture of the magnitude of deficiencies that might have occurred. The findings show that there is a tendency to prefer even numbers and avoid odd numbers in census enumeration in Lesotho. Also, the Whipple's index declined from 115 in 1986 to 106 in 1996, while the Myers' and Bachi's indexes respectively declined from 11 and 7 in 1986 to 9 and 6. These results suggest that there are deficiencies in the Lesotho data. A closer assessment of the data is undertaken by curtailing part of the erratic fluctuations in single year age distribution via grouping the data in quinary ages and applying the age–sex accuracy index. The index yields a value of 35 for the 1986 census and 32 for the 1996 census. The findings show that, though there are some distortions in the reported age–sex distributions of Lesotho, the data are fairly accurate and point to a modest improvement in quality over the decade.


(Journal of Social Development in Africa: 2003 18 (1): 111-128)