This study documents a marked discrepancy between the nutritional status of children aged 0-5 years in Botswana when measured by national surveys compared to clinic-based surveillance. We compared the average prevalence of underweight (weight-for-age z-scores below 2 standard deviations of the mean of the Center for Disease Control (CDC)/WHO reference standards) in children 0-5 years of age. According to clinic surveillance, prevalence of underweight has fallen from 14.6 ± 0.03 to 3.5 ± 0.04 per cent between 1993 and 2010. In national surveys, it had fallen from 14.6 ± 0.01 to 11.5 ± 0.01 per cent between 1993 and 2007. We explored several possibilities to explain this discrepancy, and conclude that it is because of sampling bias in the clinic surveillance. This finding underlines the need for properly conducted surveys to ensure accurate information about the nutritional status of children.