When the magnetic field of a planet is due to self-exciting hydromagnetic dynamo action in an electrically conducting fluid core surrounded by a poorly-conducting mantle, the size of the core can be estimated from secular changes of the external magnetic observations In the accessible region above the surface of the planet. To a good approximation, the rate of change with respect to time of the magnetic flux should be very small at the bounding surface of the liquid core. The ‘magnetic” radius rc of the Earth’s liquid metallic core has been calculated by using the IGRF models for the period 1900-2000, and compared with the ‘seismic’ value rc = 3485 km of the mean core radius. Physically plausible values of rc are obtained when terms up to the tenth degree in the spherical harmonic expansion of the magnetic potential are included in the analysis. Typical values of the fractional error (rc–rc)/rc amount to between -0.08 and 0.21. This error is particularly low for the DGRF 1980 but, somewhat surprisingly, apparently fluctuates with the time epoch of the IGRF model. The main source of error of the determination of the magnetic radius of the Earth’s core appears to be the incomplete knowledge of the secular variation of the geomagnetic field. The irregular behavior of the geomagnetic field models during the 1945-1955 epoch regarding the resolution of the core radius, magnetic flux and spectrum is investigated. These results provide observational support for theoretical models of the geomagnetic secular variation that treat the Earth’s core as a nearly perfect electric conductor on decade time-scales. In principle, the method can also be applied to other planets when magnetic measurements of sufficient accuracy and detail and of their secular changes become available with orbiting space-probes. The errors indicated in this approach would be acceptable if no other means were available to estimate the core radius of a planet.