Augmenting census and sample records for estimating information communications technology (ICT) penetration rates at the lowest administrative level using an interdisciplinary approach

Abstract

This paper expounds an interdisciplinary statistical technique and segmentation approach in augmenting census and sample survey records, aimed at developing timely information communications technology (ICT) penetration rates at the lowest administrative level (LAL). The Malaysian experience, in particular results obtained in augmenting the Population and Housing Census 2000 and the Household Basic Amenities and Income Survey (HBAIS) 2004 records are briefly discussed. The methodology considers both design based and model based concepts and approaches that are typically deployed in Small Area Estimation (SAE) procedures. It is proposed that census and survey records at LAL be augmented through two parts. The interdisciplinary statistical procedures considered include a design based approach for $n⩾$ 400, which is adequately large enough for computing rates at LAL and a model based approach for other categories. In the model based approaches, computation of a geometric mean (GM) between census and sample rates for 30 $⩽n<$ 400 category; clustering of LAL to form strata for $n<$ 30 category before GM techniques are deployed; “borrowing strengths” from neighbouring LAL for category $n=$ 0 and finally, assuming no change in the rates for cases $N=$ 0, where there is hardly any ICT penetrations during the census year. The interdisciplinary statistical procedures considered include design based and model based approaches for SAE procedures; segmentation approach; geometric mean (GM) procedures; Goal-Post Method for converting rates into indexes; and Dalenius-Hodges Method of Stratification technique for classifying the 927 LAL under 8-S Framework that provides ease of understanding of the results in public policy context.