All absolute paleointensity data published in peer-reviewed journals were recently compiled in the GEOMAGIA50 database. Based on the information in GEOMAGIA50, we reconstruct variations in the geomagnetic dipole moment over the past 50 kyr, with a focus on the Holocene period. A running-window approach is used to determine the axial dipole moment that provides the optimal least-squares fit to the paleointensity data, whereas associated error estimates are constrained using a bootstrap procedure. We subsequently compare the reconstruction from this study with previous reconstructions of the geomagnetic dipole moment, including those based on cosmogenic radionuclides (10Be and 14C). This comparison generally lends support to the axial dipole moments obtained in this study. Our reconstruction shows that the evolution of the dipole moment was highly dynamic, and the recently observed rates of change (5% per century) do not appear unique. We observe no apparent link between the occurrence of archeomagnetic jerks and changes in the geomagnetic dipole moment, suggesting that archeomagnetic jerks most likely represent drastic changes in the orientation of the geomagnetic dipole axis or periods characterized by large secular variation of the non-dipole field. This study also shows that the Holocene geomagnetic dipole moment was high compared to that of the preceding ∼ 40 kyr, and that ∼ 4 · 10^22 Am^2 appears to represent a critical threshold below which geomagnetic excursions and reversals occur.