Subjective well-being measures are increasingly applied in quantitative economic analyses intended to elicit non-monetary wellbeing of individuals. However, the subjective nature of this evaluation means that measurement and comparison may be confounded by differences in context or may be sensitive to the implementation modality. We use two rounds of a large-scale panel phone survey data from Myanmar to explore whether the randomized placement of a happiness module – either at the beginning or at the end of the survey – affects respondents’ answers. Respondents who were asked the happiness module at the end are more likely to be happy – an increase of 7 percentage points – compared to those who are asked at the beginning of the survey. This result is consistent using different models and robust to inclusion of enumerator fixed effects and other enumerator and survey characteristics. A related question on worry in the same module yields similar findings. Results also sustain over the two rounds of survey in which we conducted the experiment.