Online Surveys in Latin America

Abstract

Online surveys of public opinion are less expensive and faster to administer than other surveys. However, nonprobability online samples diverge from the gold standard of probabilistic sampling. Although scholars have examined the quality of nonprobability samples in the United States and Europe, we know little about how these samples perform in developing contexts. We use nine online surveys fielded in six Latin American countries to examine the bias in these samples. We also ask whether two common tools that researchers use to mitigate sample bias-post-stratification and sample matching-improve these online samples. We find that online samples in the region exhibit high levels of bias, even in countries where Internet access is widespread. We also find that post-stratification does little to improve sample quality; sample matching outperforms the provider's standard approach, but the gains are substantively small. This is partly because unequal Internet access and lack of investment in panel recruitment means that providers are unlikely to have enough panelists in lower socioeconomic categories to draw representative online samples, regardless of the sampling method. Researchers who want to draw conclusions about the attitudes or behaviors of the public as a whole in contexts like Latin America still need probability samples.