Survey Design and Data Collection
Development economics has expanded in recent years in part because many researchers are collecting new data, driven by both policy questions and curiosity about human behavior. However, designing and implementing a new survey can be a labor intensive process, particularly when best practices are unknown -- and largely unexplored -- for measurement and survey design. This course presents advancements in many areas important to measuring and evaluating enterprise-based solutions to poverty, as discussed at a December 2009 conference, "Survey Design and Measurement in Development Economics."
Preliminary drafts of papers on these topics can be found at the conference website, http://econ.worldbank.org/conferences/surveydesign. Selected, peer-reviewed papers will be published in a forthcoming special issue of the Journal of Development Economics.
Below are selected PowerPoint presentations covering topics related to measurement. Findings are preliminary; please do not quote or cite without explicit permission from the authors.
David McKenzie, Marcel Fafchamps, Simon Quinn, and Christopher Woodruff investigate the accuracy of Personal Digital Assistants (PDAs) in measuring profits and sales, data that is integral in understanding the success of policies designed to raise the incomes of the self-employed. They find that the overall benefits of PDAs are fairly modest -- results suggest that much of the variation in microenterprise data may be geniune.
In a unique study of the sensitivity of household enterprise income and asset definitions, Krislert Samphantharak and Robert Townsend measure ROA of household enterprise under different definitions of the enterprise's income and assets. The two conclude that measurement choices depend on the particular object of a survey, but that a clear conceptual framework and well-defined variables are crucial.
Recall bias is thought to affect the quality and consistency of data. In this research, Kathleen Beegle, Calogero Carletto and Kristen Himelein evaluate recall decay in agricultural data; they generally find little evidence for recall bias, even when surveys were given months after harvest.
Researchers Li-Wei Chao, Jere Behrman, Peter Fleming, Helena Szrek, Nuno Sousa Pereira, Karl Peltzer, and Shandir Ramlagan seek out best practice methods for sampling micro and small enterprises in South Africa. They work to overcome challenges such as the lack of documentation on enterprise. After evaluating several sampling methods, the authors' preliminary conclusions suggest that a stratified quota method may be successful in capturing a representative sample. It also is worth nothing that the authors utilized PDA technology in administering the survey questionnaires.
Given the large variation in household definitions used, Lori Beaman and Andrew Dillon examine how these variations affect household size, household consumption, and assets, consumption, and production statistics. The two conclude that while different research goals may require different household definitions, national surveys and in evaluations require consistency over time in definitions used for accurate measurement of variables.