Variability is an important characteristic of population dynamics, but the length of the time series required to estimate population variability is poorly understood. To this end, population variability of Macrosiphum euphorbiae (Thomas), Myzus persicae (Sulzer), and Aphis nasturtii (Kaltenbach) (Hemiptera: Aphididae) was investigated. Population variability (measured as PV, a proportion between 0 and 1) was estimated for time series of 3-62 years, giving replicate estimates for time series of 3-20 years that were normally distributed. Mean values for PV were more uniform for a time series of 12 years or longer than for shorter ones. The standard deviation of PV declined to a minimum at 12-15 years, as the length of the time series increased. Discrimination of estimates of PV was reliable for 15-year time series and longer, but not necessarily for shorter ones. Although M. euphorbiae had a relatively low PV, the coefficient of variation of that PV (12.5), was higher than for the other two species (3.5, 4.5). For robust estimates of PV, a time series of 15 years is recommended, because it minimises the standard deviation of PV, and discriminates values of PV that differ by 0.06 on a 0-1 scale.