Own-children fertility estimates are somewhat in error when the life tables used in generating them assume constant mortality, as when reliable life tables are available for only a single date in the recent past. When they are specified by geographic subdivision and socioeconomic characteristics, they are somewhat in error when the life tables used in generating them lack geographic and socioeconomic detail. This paper attempts to assess the general magnitude of these errors, by generating and comparing several sets of own-children fertility estimates from the 1970 Census of Thailand, based on 3 alternative mortality assumptions: 1) That mortality was constant during the 15 years previous to the 1970 Census at specified levels. 2) & 3) That mortality conforms to a set of changing life tables, the estimates in these cases deriving from different calculations. To illustrate the impact of the alternative mortality assumptions, own-children fertility estimates are examined by region, rural-urban residence, and education. The authors conclude that the error in own-children fertility estimates caused by using constant mortality is small (less than 2% for Thailand). The error in regional fertility estimates is also 2% or less, and the errors in residence-specific and education-specific fertility estimates are also small (less than 5 and 8%). Unfortunately, the assumption of similar mortality across geographic subdivisions and socioeconomic characteristics results in errors of own-children fertility estimates for these factors that are not unidirectional. Given the positive association between mortality and fertility levels, the result is that fertility levels are usually underestimated. Although absolute errors in estimated fertility differentials are usually small, relative errors are frequently large.