Efficient Stratification Method for Socioeconomic Survey in Remote Areas

TL;DR

This paper proposes an efficient stratification method for socio-economic surveys in remote Indonesian areas, combining welfare and geographic difficulty strata to reduce costs and improve accuracy through simulation.

Contribution

It introduces a novel stratification approach using Polychoric PCA and simulation to optimize survey efficiency in remote regions.

Findings

Identified stratification scenarios with low cost and small variance

Demonstrated improved efficiency in Papua Province case study

Provided a framework for balancing cost and accuracy in remote surveys

Abstract

The problems that exist in implementing a sampling design for socio-economic surveys in remote areas in Indonesia are high cost of the survey, low response rate, and less accurate. Therefore, the sampling design needs to be developed, one of which is to improve the efficiency of the stratification procedure. Stratification of census block in remote areas can be developed by combining the strata of welfare concentration and the strata of geographic difficulty by simulating the various alternatives number of strata and the various alternatives sample allocation. The strata of welfare concentration and the strata of geographic difficulty are constructed by Polychoric Principal Component Analysis. The strata of welfare concentration aim to improve statistical efficiency, while the strata of geographic difficulty are used to improve cost efficiency. The estimation procedure is performed at the domain level and population level. The simulation study focus on Papua Province by using the 2010 Population Census data and the 2011 Village Potency data. Some sampling scenarios can be categorized into four quadrants, the first quadrant with small sampling variance and low cost, the second quadrant with big sampling variance and low cost, the third quadrant with big sampling variance and high cost, and the fourth quadrant with small sampling variance and high cost. Based on these simulation results, several alternative scenarios of efficient stratification with small sampling variance and low cost of the survey are obtained.