The two main data sources for the estimation of total fertility in Mongolia give figures that differ by between one-half and one child per woman. We investigated two possible explanations for these discrepancies: (i) that data on births collected in the vital registration system are incomplete and under-estimate fertility; and (ii) that fertility is over-estimated when estimates are based on data from sample surveys. We find that the under-counting of single women is the main source of the discrepancies. After adjustment, survey-based fertility estimates are very close to the figures based on data from the vital registration system. The analysis underscores the risks of taking survey-based demographic indicators as the 'Gold Standard'. Demographers need to undertake more systematic studies on this important issue.