Alternatives to full listing for second stage sampling Methods and implications
International best practice on survey design recommends using a complete listing to develop the second-stage sampling frame for a household survey. In certain contexts, though, this approach may not be practical or possible due to concerns ranging from cost to security. This paper focuses specifically on security constraints encountered during the planning of a survey in Mogadishu, Somalia. The paper develops an artificial census for three neighborhoods based on data from the Mogadishu High Frequency Survey. Simulation models are then used to select individual dwellings to examine the implications of the choice of second stage selection methodology on bias and variance. Among the other findings, the simulations show that the bias introduced by a random walk design leads to the underestimation of the number of households in poverty by approximately 10 percent, with unweighted random point selection design leading to overestimation of a similar magnitude. The paper also discusses the time required and technical complexity of the associated back-office preparation work and weight calculations for each method. The paper concludes by considering practicality of each method, including the ease of implementation and options for remote verification, and outlines areas for future research and pilot testing.
Himelein, K., Eckman, S., Murray, S., & Bauer, J. (2017). Alternatives to full listing for second stage sampling: Methods and implications. Statistical Journal of the IAOS, 33(3), 701-718. https://doi.org/10.3233/SJI-160341