Relevance&Research Question: The increasing popularity of web-surveys has triggered a heated debate about their quality for scientific use. Web-surveys offer advantages such as worldwide coverage, cost benefits and fast data collection, but are mostly not representative. Findings for developed countries consistently show that young highly educated men are overrepresented in web-surveys. While the topic of representativeness has been extensively discussed for developed countries, the question remains whether similar biases can be found for developing countries. This paper aims to explore to what extent web and face-to-face surveys produce reliable data regarding socio-demographic characteristics for eight developing countries (China, Ghana, Indonesia, Kenya, Mozambique, Senegal, Uganda).
Methods&Data: We pooled data of the self-administered WageIndicator web-survey and the representative WageIndicator face-to-face surveys on work and wages (8 countries, 2010-2012, N = 16,026 web-sample; N = 18,392 face-to-face sample). Using logit analysis, the chance of inclusion in the web-sample is explored, testing three hypotheses. Comparing the web and face-to-face samples with population characteristics, we assume that both modes are confronted with the same hard-to-reach groups (H1). We assume that young, high educated, male, single and high income individuals are overrepresented in the web-sample (H2). We assume that these effects will be smaller in countries with a higher internet penetration rate (H3).
Results: Whereas no support is found for hard-to-reach groups (H1), the analysis shows evidence for good-to-reach groups: men and women aged 20-35 are highly overrepresented in both modes. Concerning H2 across all countries we find support that individuals aged 20-24, male, single, high educated and with a high income are more likely to be included in the web-survey. H3 is not supported. In contrast, we find smaller effects of socio-demographic characteristics (except gender) on web-survey participation for low access countries.
Added Value: Considering the problems related to surveying in developing countries web-surveys might be a promising cost- and time-efficient data collection tool to access data of populations so far under researched. This study contributes to the understanding of the nature of bias in volunteer web-surveys. Though Internet access is still low in developing countries, their fast increase makes a growth in web-surveys likely.