Page 247 - Special Topic Session (STS) - Volume 4

STS583 Michael W. et al.
Addressing the issue of missing or non-ideal
sampling frames in household surveys in
developing countries through remote sensing
data
Michael Wild; Brian Blankespoor; Siobhan Murray; Talip Kilic
The World Bank

Abstract
Household surveys are the most important data source on the socio-economic
conditions of the population living in Low- (LIC) and Middle-Income (MIC)
countries. In most cases the surveys are design based probabilities surveys
requiring a sampling frame. However, in many cases, the existing frames do
not fulfill the basic requirements of an ideal frame, namely completeness,
currentness and informativeness. Surveys based on an inadequate sampling
frame may deliver imprecise or biased estimates. Since any information related
to sampling errors is based on this frame, the errors related to inadequate
sampling frames usually remain undiscovered. The problem of inadequate
sampling frame is quite common. High Income Countries (HIC) have an
abundance of administrative data to address these problems. However, LIC
and MIC countries most likely do not have sufficient high quality
administrative data and therefore exclusively rely on their decennial census.
To address this considerable limitation, we compare a sample drawn from a
census based sampling frame to samples using stratification from satellite data
including landcover, gridded population data or share of built-up area for a
single province in Malawi. We show that the deviation between the estimates
and the true value are within the expected interval for all designs and frames,
and that the sampling frame data derived from satellite data, performs either
as good, or in some cases even outperforms the pure census-based frame
results. Our research therefore provides evidence to support the use of
satellite data in the construction of household sampling frames either in
combination with census data or even as a stand-alone solution.

Keywords
Sampling Methods, Census, Remote Sensing

1. Introduction
Household surveys are the most important data source on the socio-
economic conditions of the population living in Low- (LIC) and Middle-Income
(MIC) countries. And in most cases these surveys are design-based probability
surveys. One important prerequisite for this type of surveys is a sampling
frame.
However, in many cases, these frames do not fulfill the basic requirements
of an ideal sampling frame, namely completeness, currentness and

236 | I S I W S C 2 0 1 9