Authors
by Muñoz J, Langeraar W.
Abstract
Most national household surveys visit a sample of from a few to several thousands of households,
selected with known (albeit often unequal) probabilities from among all households in the country. The
total sample is typically allocated into geographical strata in order to deliver estimates for the major
administrative subdivisions of the country. Within each stratum, the sample is selected in two stages,
with small area units of a few hundred households (called Primary Sampling Units, or PSUs) in the first
stage, and individual households in the second stage. The first-stage sample frame (i.e., the list of all
PSUs in the country) is based on the information collected by the most recent census, and thus tends to
be common to many surveys, whereas the second-stage frame tends to be specifically developed for
each survey, by way of a household listing operation conducted in the selected PSUs. PSUs are typically
selected with probability proportional to size (the so called pps technique), and then a fixed number of
households (called a cluster) is selected with equal probability in each selected PSU.
This sampling strategy – hereafter referred to as “the standard strategy” – is critically dependent on
reliable and accessible census databases and census cartography. National statistical agencies are the
sole source of such data in many countries, which can make them hard to reach. Even when available,
census databases are often unreliable and almost always outdated. The conditions are particularly acute
in Myanmar, where a reliable census has not been conducted for decades and household surveys are
urgently required.
In this note we propose an alternative strategy, intended to deliver a sample of households with welldefined probabilities, on the basis of Geographic Information System (GIS) techniques, GPS, highresolution satellite imagery, and additional information sources other than censuses, in particular the
LandScan population database.
The following GIS software packages have been used in the development of this alternative strategy:
Google® Earth Prov. 7.1.1.1888
Myanmar sampling strategy Muñoz & Langeraar – November 2013
Page 3
ESRI® ArcGIS v. 10.1 advanced (formerly ArcInfo)
Blue Marble® Global Mapper v. 15
We specifically focus on Myanmar, but the techniques can be adapted to other countries.