How Well Do Internet-Based Surveys Track Labor Market Indicators in Middle-IncomeCountries ?

Online surveys are convenient, cost-effective, speedy, and increasingly popular instrumentsfor data collection. This study investigates whether online surveys that used Random Domain Intercept Technology torecruit respondents were accurately measured labor market outcomes in six middle-income countries in the aftermath ofthe COVID-19 pandemic. Compared with the national average, online surveys oversampled males, youth, those with higherlevels of education, and those in smaller households. Reweighting using propensity score estimates fails toequalize the means of variables excluded from the model. When comparing the employment-to-population ratio from theinternet surveys to the most recent relevant nationally representative surveys, the average deviation is 30 percent.Reweighting using propensity scores in that case worsened the bias. Internet survey estimates of informal andself-employment rates also tend to be inconsistent with benchmark data, although the latter are available for fewercountries. Overall, the results suggest that despite the advantages and convenience of recruiting internet surveyparticipants through Random Domain Intercept Technology, the resulting sample is not representative and even afterpropensity score reweighting, it can yield estimates that are at odds with nationally representative surveys.