The effective population size (Ne) is a foundational parameter in population genetics, yet it is typically treated as a constant when modeling recent human evolution. Using 17,629 ancient European genomes from the Allen Ancient DNA Resource spanning 0-8000 years before present, we measured drift variance at approximately 890,000 putatively neutral loci across 500-year time bins. Drift variance, corrected for sampling error, varied 3.3-fold between Early Neolithic (maximum) and Medieval (minimum) periods in the pan-European dataset. To control for population structure between geographically distinct samples, we repeated the analysis using only British and Irish samples (n=1,388). The single-population analysis yielded a 4.6-fold variation in drift variance, indicating that the pan-European estimate was conservative. These results demonstrate that Ne cannot be treated as a time-invariant parameter across the Holocene. Calculations of selection thresholds, fixation probabilities, and expected allele frequency trajectories that assume constant Ne are systematically biased, with the direction and magnitude of bias depending on the time period under consideration.