Can phone surveys be representative in low- and middle-income countries? An application to Myanmar

Abstract

For decades, in-person data collection has been the standard modality for nationally and sub-nationally representative socio-economic survey data in low- and middle-income countries. As the COVID-19 pandemic rendered in-person surveys impossible and unethical, the urgent need for rapid monitoring necessitated researchers and statistical agencies to turn to phone surveys. However, apart from pandemic-related factors, a variety of other reasons can render large segments of a population inaccessible for in-person surveys, including political instability, climatic shocks, and remoteness. Such circumstances currently prevail in Myanmar, a country facing civil conflict and political instability since the February 2021 military takeover. Moreover, Myanmar routinely experiences extreme weather events and is characterized by numerous inaccessible and remote regions due to its mountainous geography. We describe a novel approach to sample design and statistical weighting that has been successfully applied in Myanmar to obtain nationally and sub-nationally representative phone survey data. We use quota sampling and entropy weighting to obtain a better geographical distribution compared to recent in-person survey efforts, including reaching respondents in areas of active conflict. Moreover, we minimize biases towards certain household and respondent characteristics that are usually present in phone surveys, for example towards well-educated or wealthy households, or towards men or household heads as respondents. Finally, due to the rapidly changing political and economic situation in Myanmar in 2022, the need for frequent and swift monitoring was critical. We carried out our phone survey over four quarters in 2022, interviewing more than 12,000 respondents in less than three months each survey. A survey of this scale and pace, though generally of much shorter duration than in-person interviews, could only be possible on the phone. Our study proves the feasibility of collecting nationally and sub nationally representative phone survey data using a non-representative sample frame, which is critical for rapid monitoring in any volatile economy.