The Sustainable Development Goal drinking water indicators include microbiological safety measures, whereas the Millennium Development Goal indicator ‘improved water’ may be microbiologically unsafe. In rural Vanuatu, we undertook household surveys, child anthropometry, and tested stored drinking water, to investigate relationships between water and undernutrition. Using Escherichia coli most probable number, we categorized results according to Compartment Bag Test drinking water cutoffs: <1/100 mL (safe), 1–10/100 mL (intermediate risk), >10–100/100 mL (high risk), and >100/100 mL (very high risk). Of 201 households, 191 (95%) had microbiologically unsafe drinking water, regardless of ‘improved’ status. We investigated cross-sectional associations between households with microbiologically safer drinking water (≤10 E. coli/100 mL) versus ‘improved water’ and undernutrition among children. Of children under 5, 145 (48.8%, 95% CI: 42.8, 54.8) were stunted and 59 (19.1%, 95% CI: 14.4, 23.8) were underweight. Among households with ‘improved water’, the adjusted prevalence ratio (95% CI) of stunting was 0.61 (0.46, 0.80) and underweight was 0.46 (0.29, 0.73) compared with ‘unimproved water’. However, we found no association between having drinking water with ≤10 E. coli/100 mL at one point in time and undernutrition. Longer-term variations in water quality and unmeasured conditions beyond water may have contributed to these associations.